quantum singular complete integrability · initiated in classical mechanics in [de, ho] and uses...

QUANTUM SINGULAR COMPLETE INTEGRABILITY

THIERRY PAUL AND LAURENT STOLOVITCH

Abstract. We consider some perturbations of a family of pairwise commuting linear quantumHamiltonians on the torus with possibly dense pure point spectra. We prove that the Rayleigh-Schrodinger perturbation series converge near each unperturbed eigenvalue under the form ofa convergent quantum Birkhoff normal form. Moreover the family is jointly diagonalised by acommon unitary operator explicitly constructed by a Newton type algorithm. This leads to thefact that the spectra of the family remain pure point. The results are uniform in the Planckconstant near ~ 0. The unperturbed frequencies satisfy a small divisors condition and weexplicitly estimate how this condition can be released when the family tends to the unperturbedone. In the case where the number of operators is equal to the number of degrees of freedom - i.e.full integrability - our construction provides convergent normal forms for general perturbationsof linear systems.

Contents

1. Introduction 2Notations 72. Strategy of the proofs 73. The cohomological equation: the formal construction 93.1. First order 103.2. Higher orders 113.3. Toward estimating 114. Norms 125. Weyl quantization and first estimates 145.1. Weyl quantization, matrix elements and first estimates 145.2. Fundamental estimates 166. Fundamental iterative estimates: Brjuno condition case 227. Fundamental iterative estimates: Diophantine condition case 298. Strategy of the KAM iteration 309. Proof of the convergence of the KAM iteration 319.1. Convergence of the KAM iteration I: constraints on ω 319.2. Convergence of the KAM iteration II: general ω 369.3. Proof of Theorem 29 379.4. Convergence of the KAM iteration III: the classical limit 429.5. Convergence of the KAM iteration IV: the Diophantine case 479.6. Bound on the Brjuno constant insuring integrability 4710. The case m l 48References 48

Research of L. Stolovitch was supported by ANR grant “ANR-10-BLAN 0102” for the project DynPDE .

1

2 THIERRY PAUL AND LAURENT STOLOVITCH

1. Introduction

Perturbation theory belongs to the history of quantum mechanics, and even to its pre-history, as it was used before the works of Heisenberg and Schrodinger in 1925/1926. Thegoal at that time was to understand what should be the Bohr-Sommerfeld quantum condi-tions for systems nearly integrable [MB], by quantizing the perturbation series provided bycelestial mechanics [HP]. After (or rather during its establishment) the functional analysispoint of view was settled for quantum mechanics, the “modern” perturbation theory tookplace, mostly by using the Neumann expansion of the perturbed resolvent, providing effi-cient and rigorous ways of establishing the validity of the Rayleigh-Schrodinger expansionand leading to great success of this method, in particular the convergence under a simpleargument of size of the perturbation in the topology of operators on Hilbert spaces [TK],and Borel summability for (some) unbounded perturbations [GG, BS]. On the other hand,by relying on the comparison between the size of the perturbation and the distance betweenconsecutive unperturbed eigenvalues, the method has two inconveniences: it remains localin the spectrum in the (usual in dimension larger than one) case of spectra accumulatingat infinity and is even inefficient in the case of dense point unperturbed spectra which canbe the case in the present article.

In the present article, we consider some commuting families of operators on L2pTdq closeto a commuting family of unperturbed Hamiltonians whose spectra are pure point andmight be dense for all values of ~. As already emphasized, standard (Neumann seriesexpansion) perturbation theory does not apply in this context. Nevertheless, we prove thatthe pure point property is preserved and moreover, we show that the perturbed spectra areanalytic functions of the unperturbed ones. All these results are obtained using a methodinspired by classical local dynamics, namely the analysis of quantum Birkhoff forms. Letus first recall some known fact of (classical) Birkhoff normal forms.

In the framework of (classical) local dynamics, Russman proved in [Ru] (see also [Bru])the remarkable result which says that, when the Birkhoff normal form (BNF), at anyorder, depends only on the unperturbed Hamiltonian, then it converges provided that thesmall divisors of the unperturbed Hamiltonian do not accumulate the origin too fast (werefer to [Ar2] for an introduction to this subject). This leads to the integrability of theperturbed system. On the other hand, Vey proved two theorems about the holomorphicnormalization of families of l1 (resp. l) of commuting germs of holomorphic vector fields,volume preserving (resp. Hamiltonian) in a neighborhood of the origin of Cl (resp. C2l)(and vanishing at the origin) with diagonal and independent 1-jets [JV1, JV2].

These results were extended by one of us in [LS1, LS2], in the framework of general localdynamics of a families of of 1 ¤ m ¤ l commuting germs of holomorphic vector fields neara fixed point. It is proved that under an assumption on the formal (Poincare) normal formof the family and and under a generalized Brjuno type condition of the family of linearparts, there exists an holomorphic transformation of the family to a normal form. Thisfills up therefore the gap between Russman-Brjuno and the complete integrability of Vey.

QUANTUM SINGULAR COMPLETE INTEGRABILITY 3

In these directions, we should also mention works by H. Ito [It] and N.-T. Zung [Zu] inthe analytic case and H. Eliasson [El] in the smooth case, and Kuksin-Perelman [KP] for aspecific infinite dimensional version.

In [GP] one of us (the other) gave with S. Graffi a quantum version of the Russmanntheorem in the framework of perturbation theory of the quantization of linear vector fieldson the torus Tl. Moreover, in this setting, it is possible to read on the original perturbationif the Russman condition is satisfied and the results are uniform in the Planck constantbelonging to r0, 1s. The method seats in the framework of Lie method perturbation theoryinitiated in classical mechanics in [De, Ho] and uses the quantum setting established in[BGP].

The goal of the present paper is to provide a full spectral resolution for certain familiesof commuting quantum Hamiltonians, not treatable by standard methods due to possi-ble spectral accumulation, through the convergence of quantum normal Birkhoff formsand underlying unitary transformations. These families generalize the quantum versionof Rusmman theorem treated in [GP], to the quantum version of “singular complete inte-grability” treated in [LS1]. The methods use the quantum version of the Lie perturbativealgorithm together with a newton type scheme in order to overcome the difficulty createdby small divisors.

Let m ¤ l P N. For ω pωiqi1...m with ωi pωji qj1...l P Rl, let us denote byLω pLωiqi1...m, the operator valued vector of components

Lωi i~ωi.∇x i~l

j1

ωjiBBxj , i 1 . . .m

on L2pTlq.We define the operator valued vector H pHiqi1...m by

H Lω V, (1.1)

where V is a bounded operator valued vector on L2pTlq whose action is defined after afunction V : px, ξ, ~q P T Tl r0, 1s ÞÑ Vpx, ξ, ~q P Rm by the formula (Weyl quantization)

pV fqpxq »RlRl

Vppx yq2, ξ, ~qei ξpxyq~ fpyq dydξp2π~ql , (1.2)

where in the integral fpq and Vppxq2, ξ, ~q are extended to Rl by periodicity (see Section5.1 for details). We make the following assumptions.

Main assumptions

(A1) The family of frequencies vectors ω fulfills the generalized Brjuno condition


8

l1

logM2k

2k 8 where MM : min

1¤i¤mmax

0|q|¤M|xωi, qy|1. (1.3)

We will sometimes impose to ω tu fulfill the strongest collective Diophantinecondition: there exist γ ¡ 0, τ ¥ l such that

@q P Zl, q 0, min1¤i¤m

|xωi, qy|1 ¤ γ|q|τ . (1.4)

Remark : usually, 1MM

is denoted by ωM in the literature [Bru, LS1]

(A2) V takes the form, for some V 1 : pΞ, x, ~q P Rm Tl r0, 1s ÞÑ V 1pΞ, x, ~q P Rm,analytic in pΞ, xq and kth times differentiable in ~,

Vpx, ξ, ~q V 1pω1.ξ, . . . , ωm.ξ, x, ~q, (1.5)

(A3) The family H satisfies

rHi, Hjs 0, 1 ¤ i, j ¤ m, 0 ¤ ~ ¤ 1. (1.6)

Moreover we will suppose that the vectors ωj, j 1 . . .m are independent over R and wedefine

ω :m

j1

|ωj| m

j1

l

i1

pωijq212

(1.7)

Let us define for ρ ¡ 0, k P t0uYN and V 1 : pΞ, x, ~q P RmTlr0, 1s ÞÑ V 1pΞ, x, ~q P Rm

V 1ρ,ω,k m

j1

k

r0

Br~ V 1jL1

ρ,ω,rpRmZlqbL8pr0,1sq and ∇V 1ρ,ω,k maxi1...l

m

j1

k

r0

Br~BΞjV 1iρ,ω,k,

where denotes the Fourier transform on SpRm Tlq and L1ρ,ω,kpRm Zlq is the L1 space

equipped with the weighted norm°qPZl

³Rm|fpp, qq|p1 |ω p| |q|q r2 eρpω|p||q|qdp (See Section

4).Let us remark that V 1ρ,ω,k 8 implies that V 1 is analytic in a complex strip =x ρ,

=ξ ρω and k-times differentiable in ~ P r0, 1s.

We will denote V 1pΞq : 1p2πql

³Tl V

1pΞ, xqdx.

Our assumptions are shown to be non empty in Remark 5 and the relevance of assumption(A2) is discussed in Remark 6, both at the end of Section 2 below.


Our main result reads (see Theorems 29, 30 and 39 for more precise and explicit state-ments):

Theorem 1. Let k P NYt0u and ρ ¡ 0 be fixed. Let H satisfy the Main Assumption aboveand V 1ρ,ω,k,∇V 1ρ,ω,k be small enough.

Then there exists a family of vector-valued functions B~8pq, Bj~B~

8pq being holomorphicin t|=zi| ρ

2, i 1 . . .mu uniformly with respect to ~ P r0, 1s and 0 ¤ j ¤ k, such that the

family H is jointly unitary conjugated to B~8pLωq and therefore the spectrum of each Hi is

pure point and equals the set tpB~8qipω nq, n P Zlu where ω n p ωi, n ¡qi1...m.

Note that the use of Brjuno condition necessitates the intermediary result Theorem29 involving an extra condition on ω removed by a scaling argument in Theorem 30, asexplained in Section 8.

Our results being uniform in ~ we get as a partial bi-product of the preceding result thefollowing global version of [LS1]:

Theorem 2. Let ρ ¡ 0 be fixed. Let H be a family of m ¤ l Poisson commuting classicalHamiltonians pHiqi1...m on T Tl of the form H H0 V, H0px, ξq ω.ξ, ω and Vsatisfying assumption pA1q and V on the form Vpx, ξq V 1pω1.ξ, . . . , ωm.ξ, xq. Let finallyV 1ρ,ω,∇V 1ρ,ω,0 be small enough (here we consider V 1 as a function constant in ~).

Then H is (globally) symplectomorphically and holomorphically conjugated to B08pH0q.

Once again let us mention that our results are much more explicit, precise and complete(in particular concerning radii of convergence and unitary/symplectic conjugations) asexpressed in Theorems 29, 30 and 39 and Corollary 35.

Moreover it appears in the proofs that the statement in Theorem 1, as well as in Theorems29, 30 and 39 and Corollary 35, is valid for fixed value of the Planck constant ~ under theMain Assumption lowed down by restricting (1.6) to ~ fixed. More precisely under theMain Assumption with (A3) restricted to, e.g., ~ 1, the Theorem 1 is still valid byputting in the statement k 0 and ~ 1. Let us mention also that, as in the originalformulations in [Ru]-[LS1], one easily sees that condition (A2) can be replaced by the factthat the quantum Birkhoff normal form (see section 2 below for the precise definition) ateach order is a function of pL1, . . . , Lmq only.

Let us emphasize the two extreme cases, that is m l and m 1.

Corollary 1 (Quantum Vey theorem). Assume that the ωj P Rl, j 1, . . . , l, are inde-pendent over R. Assume that the Hi Lωi Vi, i 1, . . . , l are pairwise commuting. Letthe perturbation Vi be the quantization of any small enough analytic function Vi. Then thefamily H is jointly unitary conjugated to B~

8pLωq as defined in theorem 1.

We emphasize that this last result do not require neither a small divisors condition nor acondition on the perturbation, see Section 10. This correspond to full quantum integrability.Quantum integrability is a huge subject - see the seminal articles [CdV1, CdV2] to quoteonly two. The difference that provides our construction is the fact that our results givesconvergent result even at ~ 1 is the case of perturbations of linear systems.

Corollary 2 (consolidated Graffi-Paul theorem). Assume that ω P Rl satisfies Brjunocondition (m 1). Assume that H Lω V , where the perturbation V is small enough


and Vpξ, xq V 1pω.ξ, xq. Then H is unitary conjugated to B~8pLωq as defined in theorem

1.

The main difference between this last result and the main result of [GP] is the smalldivisors condition used (a Siegel type condition with constraints).

Le us finally mention a by-product of our resul, a kind of inverse result, obtained thanksto the fact that we carefully took care of the precise estimations and constants all a longthe proofs. This result is motivated by the remark that, though a small divisors conditionis necessary to obtain the perturbed integrability (and Brjuno condition is sufficient), sucha condition should disappear when the perturbation vanishes, as the Hamiltonian H0 isalways integrable, whatever the frequencies ω are. Our last result quantifies this remark.

Let us define, for ω satisfying (1.4) and α 2 log 2,

Bαpγ, τq 2 log2τγp τ

eαqτ

(note that Bαpγ, τq Ñ 8 as γ and/or τ Ñ 8).The next Theorem shows that, in the Diophantine case, the small divisors condition canbe released as Bαpγ, τq diverging logarithmically as the perturbation vanishes.

Theorem 3. Let k P N Y t0u and ρ ¡ 0 be fixed. Let ω and V satisfy (A1) (Diophantinecase), (A2) and (A3), and let 0 ω ¤ ω ¤ ω 8 and V 1ρ,ω,k, ∇V 1ρ,ω,k be smallenough (depending only on k).

Then there exist a constant Cω such that the conclusions of Theorems 1 hold as soonas, for some α ρ2, α 2 log 2,

Bαpγ, τq 1

3log

1

V ρ,ω,k

Cω .

See Corollary 40 for details and the Remark after on the case of the Brjuno condition.Let us remark that an equivalent result for Theorem 2 is straightforwardly obtainable.

Let us finish this section by mentioning three comments and remarks concerning ourresults.

First of all, as mentioned earlier, no hypothesis on the minimal distance between twoconsecutive unperturbed eigenvalues is required in our article. More, the spectra of ourunperturbed operators Lωi might be dense for all value of ~ (actually in the Diophantinecase for m 1, l ¡ 1 they are) so the Neumann series expansion is not possible. Form ¡ 1 the non degeneracy of the unperturbed eigenvalues is not even insured by thearithmetical property of ω because it relies on the minimum over i ¤ m of the inverseof the small denominator of the vector ωi. In fact, for a resonant ωj the operator Hj

will have an eigenvalue with infinite degeneracy, so the projection of the perturbation Vjon the corresponding and infinite dimensional eigenspace, which leads to the first orderperturbation correction to the unperturbed eigenvalue, might have continuous spectrum.Nevertheless our results show that the perturbed spectra are analytic functions of thespectra of the Lωi ’s.


Secondly, because of the fact that non degeneracy of some of the unperturbed spectra isnot even guaranteed by our assumptions, the standard argument on existence of a commoneigenbasis of commuting operators with simple spectra cannot be involved here. Thisexistence is a bi-product of our results.

Finally let us mention that, as it was the case in [GP], though our hypothesis on theperturbations are restrictive, our results, compared with the usual construction of quasi-modes [Ra, CdV3, PU, Po1, Po2], have the property of being global in the spectra (fulldiagonalization), and exact (no smoothing or Op~8q remainder), together of course withsharing the property of being uniform in the Planck constant.

Let us point out that this paper has been written in order to be self-contained

Notations

Function valued vectors in Rn will be denoted in general in calligraphic style, and operatorvalued vectors by capital letters, e.g. V pVlql1...m or V pVlql1...m.

For i, j P Zn we will denote by ij or ,ij when has already an index, the matrix element

of an (vector) operator in the basis tej, ejpxq eij.xp2πq l2 , θ P Tlu, namely

Vij pVl,ijql1...m ppei, VlejqL2pTnqql1...m,

and by V the diagonal part of V :

V ij Viiδij,

together with

V p2πql»TlVdx.

We will denote by | | the Euclidean norm on Rm (or Cm), |Z|2 m°i1

|Zi|2, and by

L2pTlqÑL2pTlq the operator norm on the Hilbert space L2pTlq.Finally for ω pωi P Rlqi1...m and ξ P Rl, p P Rm, q P Zl we will denote

ω ξ p ωi, ξ ¡Rlqi1...m P Rm, (1.8)

p.ω

m

i1

piωji

j1...l

P Rl (1.9)

and

p.ω.q m

i1

l

j1

piωji qj xp ω, qyRl . (1.10)

2. Strategy of the proofs

The general idea in proving Theorem 1 will be to construct a Newton-type iterationprocedure consisting in constructing a family of unitary operators Ur such that (norms willbe defined later)

U1r pB~

r pLωq VrqUr B~r1pLωq Vr1, (2.1)

with Vr1r1 ¤ Dr1Vr2r and B~

0pLωq Lω, V0 V .


Ur will be chosen of the form

Ur eiWr~ , Wr self-adjoint. (2.2)

It is easy to realize that (2.2) implies (2.1) if Wr satisfies the (approximate) cohomologicalequation

1

i~rB~

r pLωq,Wrs Vr Dr1pLωq OpVr2rq, (2.3)

or equivalently1

i~rB~

r pLωq,Wrs V cor Dr1pLωq OpVr2

rq, (2.4)

for any V cor such that V co

r Vrr OpVr2rq.

We will solve for each r the equation (2.4) where V cor will be obtained by a suitable

“cut-off” in order to have to solve (2.4) with only small denominators of finite order (seeBrjuno condition (1.3)).

In fact we will see in Section 3 that we can find a (scalar) solution of the (vector) equation(2.4) satisfying

1

i~rB~

r pLωq,Wrs V cor B~

r1pLωq Rr, (2.5)

where Rrk1 OpVr2rq. To do this we will remark that since the components of B~

r pLωqVr commute with each other (since the ones of Lω V do) we have that

rpB~r pLωqql, pVrql1s rpB~

r pLωqql1 , pVrqls rpVrql1 , pVrqls OpV 2r q (2.6)

which is an almost compatibility condition (see Section 3 for details).Summarizing, the solution Wr of (2.4) will provide a unitary operator Ur such that (2.1)

will hold with B~r1 B~

r Dr1 and Vr1 being the sum of three terms:

V 1r1 U1

r pB~r pLωq VrqUr pB~

r pLωq Vrq 1i~rB~

r pLωq,Wrs V 2

r1 Vr V cor

V 3r1 Rr

The choice of the family of norms r will be made in order to have that

Vr1r1 V 1r1 V 2

r1 V 3r1r1 ¤ Dr1Vr2

r

with Dr satisfyingR¹r1

D2Rr

r ¤ C2R .

Hence, we have

VR1R1 ¤ pCV00q2R

,

so that VR1R1 Ñ 0 as RÑ 8 if V0 V 0 C1 and 8 exists.

Remark 4. [Propagation of assumptions (A2)-(A3)] It is clear (and it will be explicit inthe body of the proofs of the main Theorem) that Condition (A2) will be satisfied by thesolution of equations (2.3),(2.4) as soon as Vr and V co

r do. This last condition can be easilyseen to be propagated from the decomposition Vr1 V 1

r1 V 2r1 V 3

r1 given before by

considering that Ur eiWr~ by (2.2) and Wr satisfying (A2). (A3) is obviously propagated

by (2.1).


Remark 5. [Non emptiness of the hypothesis] Consider a family of operators of the formLω B~pLωq for B~ : Rm Ñ Rm with B~ρ,ω,k 8. Then for each bounded self-adjoint operator W whose Weyl symbol W satisfies (A2) and W ρ1,ω,k 8 for some

ρ1 ¡ ρ, consider the family eiW~ pLω B~pLωqqeiW~ : Lω V : pHiqi1...m. Obviously the

family pHiqi1...m satisfies (A3). By the same argument as the one in Remark 4 one seeseasily that the Weyl symbol V of V satisfies (A2) for some V 1. Finally estimates (5.11)

and (5.12) in Proposition 16 below show that the expansion eiW~ pLω B~pLωqqeiW~

LωB~pLωqrLωB~pLωq, iW~ s 12rrLωB~pLωq, iW~ s, iW~ s. . . is actually convergent. This

implies that Vρ,ω,k is bounded. Therefore the family Lω V satisfies all the assumptionsof Section 1.

Remark 6. [Relevance of assumption (A2)] Let us recall some classical facts from dynamical

systems. Let H0 n°i1

λipx2i y2

i q be a quadratic Hamiltonian on R2n. Any analytic

higher order perturbation H H0 higher order terms is formally conjugate to a formalBirkhoff normal form Hpx2

1 y21, . . . , x

2n y2

nq. Russman-Brjuno’s theorem asserts that, if

pq H F pH0q (i.e. H is a function of that peculiar linear combinationn°i1

λipx2i y2

i qand contains no other terms), for some formal power series F of one variable and if a”small divisors” condition is satisfied, then the transformation to the Birkhoff normal formis analytic in a neighborhood of the origin. Condition pq is known as Brjuno’s conditionA (cf. [Bru]). It is a sharp condition for the analycity of the transformation to Birkhoffnormal form in the following sense : if a normal form NF doesn’t satisfy it, then it ispossible to perturb H in such way that the analytic perturbation H still has NF as normalform and the transformation from H to NF is a divergent power series. In our quantumversion, we only focus on the sufficiency of the analogue condition. The linear combination°j ωjξj in our article plays the role of ”quantum analogue” of

°i λipx2

i y2i q

3. The cohomological equation: the formal construction

In this section we want to show how it is possible to construct the solution of the equation

1

i~rB~pLωq,W s V DpLωq OpV 2q, (3.1)

where we denote by Lω, ω pωi P Rlqi1...m, the operator valued vector of components(with a slight abuse of notation) Lωi i~ωi.∇x, i 1 . . .m on L2pTlq and V is a “cut-off”ed.

Vij 0 for |i j| ¡M.

We will present the strategy only in the case of the Brjuno condition, the Diophantinecase being very close.

Let us recall also that equation (3.1) is in fact a system of m equations and that it mightseem surprising at the first glance that the same W solves (3.1) for all ` 1 . . .m.


3.1. First order. At the first order the cohomological equation is

rLω` ,W si~

V` D`pLωq, l 1 . . .m (3.2)

solved on the eigenbasis of any Lω` by D`pLωq diagpV`q and

Wij pV` D`qijiω` pi jq . (3.3)

Indeed, since Lωl is selfadjoint, we have

ej, rLωl ,W sei ¡ ej, LωlWei WLωlei ¡ Lωlej,Wei ¡ ej,WLωlei ¡ iωl.pj iq ej,Wei ¡

In (3.3) we will picked up, for every ij such that |i j| ¤ M , an index ` ìj whichminimize the quantity

|xω`q , qy|1 : min1¤i¤m

|xωi, qy|1 ¤ MM . (3.4)

We define W by

Wij pVìjqijiωìj pi jq , i j 0 (3.5)

Since rH`, H`1s 0, then we have that rL`1V`s rV`1 , L`s rV`, V`1s. Therefore, evaluatingthe operators on ej and taking the scalar product with ei, leads to

ω`1 pi jqpV`qij ω` pi jqpV`1qij prV`, V`1sqij (3.6)

that ispV`qij

ω` pi jq pV`1qij

ω`1 pi jq prV`, V`1sqij

ω` pi jqω`1 pi jq(note that when ω`1 pi jq 0 on has pV`1qij prVìj ,V`1 sqij

ωìj pijq).

Let us remark that, though rV`, V`1s is quadratic in V , it has the same cut-off propertyas V , namely prV`, V`1sqij 0 if |i j| ¡M as seen clearly by (3.6).

This means that W defined by (3.5) satisfies

rLω,W si~

V DpLωq V ,

where

pV`qij prV`, Vìj sqiji~ωìj pi jq . (3.7)

Note that this construction is different from the one used in [LS1].


3.2. Higher orders. The cohomological equation at order r will follow the same way, atthe exception that Lω has to be replaced by B~

r pLωq.The corresponding cohomological equation is therefore of the form

rB~r pLωq,Wrs

i~ Vr OppVrq2q, (3.8)

equivalent toB~r p~ω iq B~

r p~ω jqi~

pWrqij pVrqij OppVrq2q. (3.9)

Lemma 7. For B~r close enough to the identity there exists a m m matrix Arpi, jq such

thatB~r p~ω iq B~

r p~ω jqi~

pI Arpi, jqqω.pi jq, (3.10)

where I is the mm identity matrix and ω.pi jq pωl.pi jqql1...m. Moreover

Arpi, jqCmÑCm ¤ ∇pB~r B~

0qpCmÑCmqbL8pRmq ¤ maxj1...m

m

i1

∇jpB~r B~

0qiL8pRmq. (3.11)

Proof. We have

B~r p~ω iq B~

r p~ω jqi~

ω.pi jq » 1

0

BtpB~

r B~0qpt~ω.i p1 tq~ω.jq dt

~

ω.pi jq » 1

0

∇pB~

r B~0qpt~ω.i p1 tq~ω.jq rω.pi jqsdt

so Arpi, jq ³1

0∇pB~

r B~0qpt~ω.i p1 tq~ω.jqdt and the first part of (3.11) follows. The

second part is a standard estimate of the operator norm.

Plugging (7.5) in (3.9) we get that W must solve

ω.pi jqWij pI Arpi, jqq1pVrqij OppVrq2q

, (3.12)

and we are reduced to the first order case with Vr Ñ V r where

V rij : pI Arpi, jqq1 pVrqij. (3.13)

3.3. Toward estimating. We will first have to estimate V r: this will be done out of itsmatrix coefficients given by (3.13) by the method developed in Section 5.1. We will estimate

pI Arpi, jqq1 V r in section 6 by using the formula pI Arpi, jqq1 8°k0

pArpi, jqqk and

a bound of the norm of pArpi, jqqkV r of the form |C|k times the norm of V r leading to a

bound of pI Arpi, jqq1 V r of the form 11|C| times the norm of V r, by summation of the

geometric series8°k0

Ck, possible at the condition that |C| 1.

We will then have to estimate W defined through

Wij pV r`ij

qijiω`ij pi jq , i j 0 (3.14)


with again pV rìj

qij 0 for |i j| ¡M . We get

|Wij| ¤ MM |pVìjqij|,and we will get an estimate of W , W ¤ MMV r, for a norm to be specified later.

Finally we will have to estimate

pV rl qij

prV rl , V

rìj

sqiji~ωìj pi jq . (3.15)

We will get immediately V rl ¤ MMP , Pij

prV rl ,V rìj sqiji~ and the estimate of the

commutator will be done by the method developed in Section 5.In the two next sections we will define the norms and the Weyl quantization procedure

used in order to precise the results of this section,

4. Norms

Let m, l be positive integers. For F P C8pRm Tl r0, 1s;Cq we will use the followingnormalization for the Fourier transform.

Definition 8 (Fourier transforms). Let p P Rm and q P Zl

Fpp, x, ~q 1

p2πqm»Rm

Fpξ, x, ~qeixp,ξy dξ (4.1)

Fpξ, q; ~q 1

p2πql»TlFpξ, x; ~qeixq,xy dx (4.2)

Fpp, q, ~q 1

p2πqml»RmTl

Fpξ, x, ~qeixp,ξyixq,xydξdx (4.3)

1

p2πqm»Rm

Fpξ, q, ~qeixp,ξydξ (4.4)

1

p2πql»TlFpp, x, ~qeixq,xydx (4.5)

Note that

Fpξ, x, ~q »Rm

Fpp, x, ~qeixp,ξy dp (4.6)

¸qPZl

Fpξ, q; ~qeixq,xy (4.7)

¸qPZl

»Rm

Fpp, q, ~qeixp,ξyixq,xy dp (4.8)

Set now for k P NY t0u and p ω p °j1...m

pj.ωijqi1...l:

µkpp, qq : p1 |p ω|2 |q|2q k2 (4.9)


(note that µrpp p1, q q1q ¤ 2k2µrpp, qqµrpp1, q1q because |xx1|2 ¤ 2p|x|2|x1|2q and that

|p ω| Ñ 8 as |p| Ñ 8 because the vectors pωiq11...l are independent over R).

Definition 9 (Norms I). For ρ ¡ 0, F P C8pRm Tl r0, 1s;Cq we introduce the weightednorms

F:ρ F:ρ,ω : max~Pr0,1s

»Rm

¸qPZl

| Fpp, q, ~q| eρpω|p||q|q dp. (4.10)

F:ρ,ω,k F:ρ,ω,k : max~Pr0,1s

k

j0

»Rm

¸qPZl

µkjpp, qqBj~| Fpp, q, ~q| eρpω|p||q|q dp. (4.11)

Note that ω is given by (1.7) and :ρ;0 :σ.

Definition 10 (Norms II). Let Oω be the set of functions F : Rl Tl r0, 1s Ñ C suchthat Fpξ, x; ~q F 1pω ξ, x, ~q for some F 1 : Rm Tl r0, 1s Ñ C. Define, for F P Oω:

Fρ,ω,k : F 1:ρ,ω,k. (4.12)

We will also need the following definition for F P Oω:

F~ρ,ω,k :k

j0

»Rm

¸qPZl

µkjpp, qqBj~|F 1pp, q, ~q| eρpω|p||q|q dp. (4.13)

Let us note that, obviously, ~ρ,ω,k ¤ ρ,ω,k.

We will need an extension of the previous definition to the vector case. Consider nowF P C8pRm Tl r0, 1s;Cmq and G P C8pRm r0, 1s;Cmq. The definition of the Fouriertransform is defined as usual, component by component.

Definition 11. [Norms III] Let F pFiqi1...m P C8pRm Tl r0, 1s;Cmq. We define

(1)

F:ρ,ω,k m

i1

Fi:ρ,ω,k (4.14)

(2) Let

Omω

F pFiqi1...m : Rm Tl r0, 1s Ñ Cm Fi P Oω, i 1 . . .m(

(4.15)

Let F P Omω . We define:

Fρ,ω,k m

i1

Fiρ,ω,k (4.16)

Let

Ommω

F pFijqi,j1...m : Rm Tl r0, 1s Ñ Cm Fij P Oω, i, j 1 . . .m(

(4.17)

Let F P Ommω . We define:

Fρ,ω,k supi1...m

¸j1...m

Fijρ,ω,k. (4.18)


(3) Finally we denote F the Weyl quantization of F recalled in Section 5 and

F ρ,ω,k Fρ,ω,k (4.19)

J :k pρ, ωq tF | F:ρ,ω,k 8u, (4.20)

J:kpρ, ωq tF |F P J :k pρ, ωqu, (4.21)

Jkpρ, ωq tF P Oω | Fρ,ω,k 8u, (4.22)

Jkpρ, ωq tF |F P Jkpρ, ωqu. (4.23)

J ~k pρ, ωq tF P Oω | F~ρ,ω,k 8u, (4.24)

J~k pρ, ωq tF |F P J ~

k pρ, ωqu. (4.25)

Jmk pρ, ωq tF P Om

ω | Fρ,ω,k 8u, (4.26)

Jmk pρ, ωq tF |F P Jmk pρ, ωqu. (4.27)

Jmmk pρ, ωq tF P Omm

ω | Fρ,ω,k 8u, (4.28)

Jmmk pρ, ωq tF |F P Jmmk pρ, ωqu (4.29)

and J @pρ, ωq J @k0pρ, ωq, J@pρ, ωq J@

k0pρ, ωq @@ P t:,m,mmu.When there will be no confusion we will forget about the subscript ω in thelabel of the norms and also denote by J @

k pρq J @k pρ, ωq.

5. Weyl quantization and first estimates

We express the definitions and results of this section in case of scalar operators andsymbols. The extension to the vector case is trivial component by component. The readeronly interested by explicit expression can skip the beginning of the next paragraph and godirectly to Definition 5.4.

5.1. Weyl quantization, matrix elements and first estimates. In this section werecall briefly the definition of the Weyl quantization of T Tl. The reader is referred to [GP]for more details (see also e.g. [Fo]).

Let us recall that the Heisenberg group over T Tl R, denoted by HlpRl Zl Rq, is(the subgroup of the standard Heisenberg group HlpRlRlRq) topologically equivalent toRlZlR with group law pu, tqpv, sq puv, ts 1

2Ωpu, vqq. Here u : pp, qq, p P Rl, q P Zl,

t P R and Ωpu, vq is the canonical 2form on Rl Zl: Ωpu, vq : xu1, v2y xv1, u2y.The unitary representations of HlpRl Zl Rq in L2pTlq are defined for any ~ 0 as

follows

pU~pp, q, tqfqpxq : ei~tixq,xy~xp.qy2fpx ~pq (5.1)

Consider now a family of smooth phase-space functions indexed by ~, Apξ, x, ~q : Rl Tl r0, 1s Ñ C, written under its Fourier representation

Apξ, x, ~q »Rl

¸qPZl

App, q; ~qeipxp.ξyxq,xyq dp (5.2)


Definition 12 (Weyl quantization I). By analogy with the usual Weyl quantization onT Rl[Fo], the (Weyl) quantization of A is the operator Ap~q defined as

Ap~q : p2πql»Rl

¸qPZl

App, q; ~qU~pp, q, 0q dp (5.3)

(note that the factor p2πql in (5.3) is due to the (convenient for us) normalization of theFourier transform in Definition 8).

It is a straightforward computation to show that, considering f P L2pTlq and Vppxq2qas periodic functions on Rl, we get the equivalent definition

Definition 13 (Weyl quantization II).

pAp~qfqpxq :»RlξRly

Appx yq2, ξ, ~qei ξpxyq~ fpyq dξdyp2π~ql (5.4)

Remark 14. The expression (13) is exactly the same as the definition of Weyl quantizationon T Rl except the fact that f is periodic. Note that Ap~qf is periodic thanks to the factthat Apx, ξ, ~q is periodic:³Appx2π yq2, ξqei ξpx2πyq

~ fpyqdξdy~l ³Appx2π y2πq2, ξqei ξpxyq~ fpy2πqdξdy~l ³

Appx yq2 2π, ξqei ξpxyq~ fpyqdξdy~l ³Appx yq2, ξqei ξpxyq~ fpyqdξdy~l pAp~qqfpxq.

The first results concerning this definition are contained in the following Proposition.

Proposition 15. Let Ap~q be defined by the expression (5.4). Then:

(1) @ρ ¡ 0, @ k ¥ 0 we have:

Ap~qBpL2pTlqq ¤ Aρ,k (5.5)

and, if Apξ, x, ~q A1pω ξ, x; ~qAp~qBpL2pTlqq ¤ A1ρ,k. (5.6)

(2) Let, for n P Zl, enpxq einx

p2πql . Then for all m,n in Zl,

xem, Ap~qenyL2pTlq Appm nq~2,m n, ~q (5.7)

(3) Reciprocally, let Ap~q be an operator whose matrix elements satisfy (5.7) for someA belonging to J @,@ P t:,m,mmu. Then Ap~q is the Weyl quantization of A.

Proof. (5.7) is obtained by a simple computation. It also implies that

Ap~qem2L2pRlq

¸qPZl

|Ap~pm qq2,m q, ~q|2 ¤ supξPRl

¸qPZl

|Apξ, q, ~q|2.

So that

Ap~q¸Zl

cmem2L2pRlq ¤

¸Zl

|cm|2 supξPRl

¸qPZl

|Apξ, q, ~q|2 ¤ p¸Zl

|cm|2q¸qPZl

supξPRl

|Apξ, q, ~q|

2

.


And therefore, since by (4.6)-(4.7)-(4.8) Apξ, q, ~q ³RlApp, q, ~qei ξ,p¡dp so that |Apξ, q, ~q|

¤ ³Rl |

App, q, ~q|dp,Ap~qBpL2pTlqq ¤

¸qPZl

supξPRl

|Apξ, q, ~q| ¤»Rl

¸qPZl

| App, q, ~q|dp ¤ Aρ,k, @ρ ¡ 0, k ¥ 0. (5.8)

In the case Apξ, x, ~q A1pω ξ, x; ~q we get, @ρ ¡ 0, k ¥ 0:

Ap~qBpL2pTlqq ¤¸qPZl

supξPRl

|Apξ, q, ~q| ¸qPZl

supY PRm

|A1pY, q, ~q| ¤»Rm

¸qPZl

|A1pp, q, ~q|dp ¤ A1ρ,k.

(3) is obvious.

5.2. Fundamental estimates. This section contains the fundamental estimates whichwill be the blocks of the estimates needed in the proofs of our main results. These primaryestimates are contained in the following Proposition. We shall omit to write the subscriptω in the norms.

Proposition 16. We have:

(1) For F,G P J1k pρq, FG P J1

k pρq and fulfills the estimate

FGρ,k ¤ pk 1q8kF ρ,k Gρ,k (5.9)

(2) There exists a positive constant C 1 such that for F P Jmk pρq and for G P J1k pρq, we

have, @δ1 ¡ 0, δ ¥ 0, ρ ¡ δ δ1,rF,Gsi~

ρδδ1,k

¤ 2pk 1q8ke2δ1pδ δ1qF ρ,kGρδ,k, (5.10)

1

d!rG, . . . rGloooomoooon

d times

, F s spi~qdρδ,k ¤ 1

2π

2p1 kq8k

δ2

d

F ρ,kGdρ,k, (5.11)

and rLω, Gsi~

ρδ,k

¤ ω

eδGρ,k (5.12)

(3) For F ,G P J 1k pρq, FG P J 1

k pρq and

FGρ,k ¤ pk 1q4kFρ,k Gρ,k. (5.13)

(4) Let V pVlql1...m P Jmk pρq and let W be defined by xem,Weny xem,Vlmneny

ω`mnpmnq where,

@m,n P Z, the index lmn is such that |xω`mn ,m ny|1 : min1¤i¤m

|xωi,m ny|1 .

Then

W ρd,k ¤ γτ τ

pedqτ V ρ,k (5.14)

in the Diophantine case and (obviously) when |xem, Vlmneny| 0 for |m n| ¡M ,

W ρ,k ¤ MMV ρ,k (5.15)

in the case of the Brjuno condition (MM defined by (1.3)).


(5) Let finally V pVlql1...m P Jmk pρq and let P be defined by pPlqij prVl,V`ij sqiji~ for any

choice of pi, jq Ñ `ij. Then P pPlql1...m P Jmk pρ δq, @δ1 ¥ 0, δ ¡ 0, ρ ¡ δ δ1

and

P ρδδ1,k ¤2pk 1q8ke2δ1pδ δ1qV ρ,kV ρδ,k (5.16)

(6) Moreover let F : ξ P Rm ÞÑ Fpξq P Rm be in Jmk pρq. Let us define ∇F the matrix

pp∇Fqijqi,j1...m with

p∇Fqij : BξiFj. (5.17)

Then, for all δ ¡ 0, ∇F P Jmmk pρ δq and

∇Fρδ,k ¤ 1

eδFρ,k. (5.18)

Let us remark that, as the proof will show, Proposition 16 remains valid when the norm ρ,k is replaced by the norm ~ρ,k

Proof. Items (1) and (2) are simple extension to the multidimensional case of the cor-responding results for m 1 proven in [GP]. For sake of completeness we give here analternative proof in the case m 1. The proof will use the three elementary inequalities,

µkpp p1, q q1q ¤ 2k2µkpp, qqµkpp1, q1q (5.19)

|pp.ω.q1 p1ω.qq2|k ¤ µkpp, qqµkpp1, q1q (5.20)Bk~ sinx~~

¤ |x|k1 (5.21)

|p ω q| ¤ ω maxj1...m

|pj||q| ¤ ω|p||q| (5.22)

where we have used the notation (1.10) and the definition (1.7).(in order to prove (5.19), (5.20), (5.21) and (5.22) just use |X X 1|2 ¤ 2p|X|2

|X 1|2q for all X,X 1 P R2l, |pp.ω.q1 p1ω.qq2|2 ¤ p|p ω|2 |q|2qp|p ω1|2 |q1|2q, sinx~~

x³0

cos ps~qds and |p ω q| ¤m°j1

|pj||l°

i1

ωijqi| m°j1

|pj||ωj q| ¤m°j1

|pj||ωj||q| by Cauchy-

Schwarz, respectively).We start with (5.9). Since F,G P J1pρq we know that there exist two functions F 1,G 1

such that the symbols of F,G are Fpξ, xq F 1pω.ξ, xq, Gpξ, xq G 1pω.ξ, xq. By (5.7) we


have that

pFGqmn ¸q1PZl

Fmq1Gq1n

¸q1PZl

Fm q1

2~,m q1

Gq1 n

2~, q1 n

¸q1PZl

Fm n q1

2~,m n q1

Gq1 2n

2~, q1

¸q1PZl

F 1ω m n q1

2~,m n q1

G 1

ω q

1 2n

2~, q1

¸q1PZl

F 1ω m n q1

2~,m n q1

G 1

ω q

1 m n pm nq2

~, q1.(5.23)

Calling P the symbol of FG we have that, by (5.7) again, pFGqmn Ppξ, qq with ξ mn2

~and q m n. Therefore

Ppξ, qq ¸q1PZl

F 1ω.ξ ω q

1

2~, q q1

G 1

ω.ξ ω q

1 q

2~, q1

, (5.24)

so we see that Ppξ, q depends only on ω.ξ: Ppξ, xq P 1pω.ξ, xq. Moreover, since by

(4.3)P 1pp, q 1

p2πqm³Rm P 1pΞ, qei Ξ,p¡dΞ we get easily by simple changes of integration

variables and the fact that the Fourier transform of a product is a convolution,

P 1pp, qq »Rm

¸q1PZl

F 1pp p1, q q1qei ~2 ppp1q.ω.q1 G 1pp1, q1qei ~2 p1.ω.pq1qqq

dp1. (5.25)

Therefore FGρ,k is equal to the maximum over ~ P r0, 1s of

k

γ0

»R2m

¸pq,q1qPZ2l

µkγpp, qq|Bγ~F 1pp p1, q q1qei ~2 pppp1q.ω.q1p1.ω.pqq1qq G 1pp1, q1q

|eρpω|p||q|qdpdp1.

(5.26)


Writing, by (5.20), that

|Bγ~F 1pp p1, q q1qei ~2 pppp1q.ω.q1p1.ω.pqq1qq G 1pp1, q1q

|

¤γ

µ0

γ

µ

γµ

ν0

γ µ

ν

|Bγµν~

F 1pp p1, q q1q||Bν~ei~2pppp1q.ω.q1p1.ω.pqq1qq||Bµ~ G 1pp1, q1q|

¤γ

µ0

γ

µ

γµ

ν0

γ µ

ν

|Bγµν~

F 1pp p1, q q1q||ppp p1q.ω.q1 p1.ω.pq q1qq2|ν |Bµ~ G 1pp1, q1q|

: PpF 1,G 1q (5.27)

¤γ

µ0

γ

µ

γµ

ν0

γ µ

ν

µνpp p1, q q1qµνpp1, q1q|Bγµν~

F 1pp p1, q q1q||Bµ~ G 1pp1, q1q|

pchanging µÑ γ1, ν Ñ ν 1 : γ γ1 νq

¤γ

γ10

γ

γ1

γγ1¸ν10

γ γ1

ν

µγγ1ν1pp p1, q q1qµγγ1ν1pp1, q1q|Bν1~ F 1pp p1, q q1q||Bγ1~ G 1pp1, q1q|

psince

m

n

¤ 2m, γ ¤ k, γ γ1 ¤ kq

¤k

γ10

2kk

ν10

2kµγγ1ν1pp p1, q q1qµγγ1ν1pp1, q1q|Bν1~ F 1pp p1, q q1q||Bγ1~ G 1pp1, q1q|

(5.28)

using (5.19) under the form

µkpp, qq ¤ 2k2µkpp p1, q q1qµkpp1, q1q

together with the fact that µkpp, qq is increasing in k and µkµk1 µkk1 .We find that

µkγpp, qqPpF 1,G 1q

¤ 4k2k2

k

γ10

k

ν10

µkγγγ1ν1pp p1, q q1qµkγγγ1ν1pp1, q1q|Bν1~ F 1pp p1, q q1q||Bγ1~ G 1pp1, q1q|

preplacing 2k2 by 2k to avoid heavy notations and since k γ1 ν 1 ¤ k γ1, k ν 1q

¤ 8kk

γ10

k

ν10

µkν1pp p1, q q1qµkγ1pp1, q1q|Bν1~ F 1pp p1, q q1q||Bγ1~ G 1pp1, q1q|. (5.29)


Note that γ disappeared from (5.29) so thek°γ0

in (5.26) gives a fractor p1kq. We get that

p1 kq18kFGρ,k is majored by the maximum over ~ P r0, 1s (note the change ν 1 Ñ γ)

¸q,q1PZl

»R2m

k

γ,γ10

µkγpp, qq|Bγ~F 1pp, qq|µkγ1pp1, q1q|Bγ1~ G 1pp1, q1q|eρpω|p|ω|p1||q||q1|qdpdp1(5.30)

which is equal to

F 1ρ,kG 1ρ,k.The proof of (5.10) follows the same lines, except that it is easy to see that, in (5.25),

ei~2pppp1q.ω.q1p1.ω.pqq1qq has to be replaced by 2 sin

~2ppp p1q.ω.q1 p1.ω.pq q1qq, since

(5.23) becomesrF,Gsi~

mn

¸q1PZl

Fmq1Gq1n Gmq1Fq1ni~

1

i~

¸q1PZl

Fm n q1

2~,m n q1

Gm n q1 pm nq

2~, q1

Gm n q1

2~,m n q1

Gm n q1 pm nq

2~, q1

(5.31)

It generates in (5.26) the replacement of |ppp p1q.ω.q1 p1.ω.pq q1qq2|νby the term

2|ppp p1q.ω.q1 p1.ω.pq q1qq2|ν1 ¤ µνpp p1, q q1qµνpp1, q1qp|p ω q1 p1 ω q|qthanks to (5.20), and we get by a discussion verbatim the same than the one contained inequations (5.27)-(5.30) that

rF,Gsi~ρδδ1,k ¤ p1 kq8k¸

q,q1PZl

»R2m

k

γ,γ10

µkγpp, qqµkγ1pp1, q1qQdpdp1, (5.32)

where thanks to (5.22),

Q |Bγ~F 1pp, qq|p|p ω q1| |p1 ωq|q|Bγ1~ G 1pp1, q1q|epρδδ1qpω|p||q|ω|p1||q1|q

¤|Bγ~F 1pp, qq||Bγ1~ G 1pp1, q1q|eρpω|p||q|qpρδqpω|p1||q1|q

pω|p||q1| ω|p1||q|qepδδ1qpω|p||q|qδ1pω|p1||q1|qq

¤ 2

e2δ1pδ δ1q |Bγ~F 1pp, qq||Bγ1~ G 1pp1, q1q|eρpω|p||q|qpρδqpω|p1||q1|q (5.33)

because (ex ¤ 1, x ¥ 0 and)

supxPR

xeαx 1

eα. (5.34)

(5.10) follows immediatly from (5.32).


The proof of (5.12) follows also the same line and is obtained thanks to the remark(5.34): indeed since rLω,W s

i~

mn

iω pm nqWmn,

we see, again by (5.7), that the symbol Qpξ, xq of rLω,W si~ is given trough the formula

Qpξ, qq prLω,W si~qmn for ξ m n

2~ and q m n.

Therefore Qpξ, qq piω qqWpξ, qq, so Qpξ, xq Q1pω.ξ, xq with

Q1pω.ξ, qq iω.q W 1pω.ξ, qq.We get immediatly Q1pp, qqepρδqpω|p||q|q ¤ ω

eδ

W 1pp, qqeρpω|p||q|q

and (5.12) follows.(5.11) is easily obtained by iteration of (5.10) and the Stirling formula: consider the

finite sequence of numbers δs dsdδ. We have δ0 δ, δd 0 and δs1 δs δ

d. Let us

define G0 : F and Gs1 : 1i~rG,Gss, for 0 ¤ s ¤ d 1. According to (5.10), we have

Gsρδds,k ¤ck

e2δdsp δdqGρ,kGs1ρδds1,k,

whereck : 2pk 1q8k.

Hence, by induction, we obtain

1

d!Gdρδ0,k ¤ cd1

k

d!e2pd1qδ0 δd2p δdqd1Gd1

ρ,k G1ρδd1,k (5.35)

¤ cdkd!e2dδ0 δd1δd1p δdqd1

Gdρ,kF ρ,k

¤ cdkd!e2dd!p δ

dqdδd1p δdqd1

Gdρ,kF ρ,k

¤ 1

2π

ckd

2

e2δ2

d1

pd 1q!d!Gdρ,kF ρ,k

1

2π

ckδ2

d?2πddded

d!

2

Gdρ,kF ρ,k

¤ 1

2π

ckδ2

dGdρF ρ,k

since d!?2πdeddd

¥ 1. This weel know inequality can be seen from Binet’s second expression

for the log Γpzq [WW][p. 251] :

log

n!

ne

n?2πn

2

» 8

0

arctanptnqe2πt 1

dt ¥ 0.


Finally (5.13) is obtained by noticing that FGρ,k has the same expression as FGρ,kafter removing the term ei

~2pp.ω.q1p1.ω.qq in (5.25).

To prove (4) it is enough to notice that by (5.7) the symbol of W satisfies W pξ, q, ~q V`q pξ,q,~qω`q q , so that

W pp, q, ~q

V 1`q pp,q,~q

ω`q q and therefore, for all r P N,

|Br~W pp, q, ~q| ¤ γ|q|τ supl1...m

|Br~ V lpp, q, ~q| ¤ γ|q|τm

l1

|Br~ V lpp, q, ~q|

out of which we deduce (5.14) by standard arguments (xτeδx ¤ p τeδqτ , x ¡ 0) in the

Diophantine case, and

|Br~W pp, q, ~q| ¤ MM supl1...m

|Br~ V lpp, q, ~q| ¤ MM

m

l1

|Br~ V lpp, q, ~q|

from which (5.15) follows.

To prove (5.18) we just notice thatBξiFjpp, q, ~q pi

Fjpp, q, ~q. So

|BξiFjpp, q, ~q| ¤ |piFjpp, q, ~q| ¤ |Fjpp, q, ~q||p|.

Therefore |Br~BξiFjpp, q, ~q|epρδq|p| ¤ 1

eδ|Br~

BξiFjpp, q, ~q|eρ|p| and (5.18) follows.(5) is an easy extension of (5.10). Indeed we find immediately, by (5.7) and the

fact that lij depends only on i j, that the Fourier transform of the symbol of Pl isPpp, q, ~q X lqpp, q, ~q whereX lq is the Fourier transform of the symbol of the oper-

ator Xlq rVl,Vlq si~ . Therefore |Br~ X lqpp, q, ~q| ¤ max

l1...m|Br~ X lpp, q, ~q|, @r ¥ 0, q P Zl. So

Plρδ,k ¤ maxl11...m

rVl, Vl1si~ρδ,k and

P ρδ,k ¤ maxl,l11...m

rVl, Vl1si~ρδ,k ¤m

l,l11

rVl, Vl1si~ρδ,k

and we conclude by using (5.10).

6. Fundamental iterative estimates: Brjuno condition case

In all this section the norm subscripts ω and k are omitted.Let us recall from Sections 2 and 3 that we want to find Wr such that

eiWr~ pHr VrqeiWr~ Hr1 Vr1 (6.1)

where Hr1 Hr hr1 and Hr BrpLωq, hr1 Vr DrpLωq and, for 0 δ ρ 8,

hr1ρ Vrρ ¤ Vrρ, Vr1ρδ ¤ DrVr2ρ, (6.2)

and that we look at Wr solving:

1

i~rHr,Wrs V co,r V co,r V r (6.3)


withV co,r Vr V Mr .

V Mr is given by

V Mrij pVrqij if |i j| ¡Mr, V Mr

ij 0 otherwise. (6.4)

(note that V co,r V r).

V r pV rl ql1...m is given by

pV rl qij

prV rl , V

rlpijqsqij

i~ωlpijq pi jq , V rij : pI Arpi, jqq1 V co

ij , V co,r Vr V Mr , (6.5)

where Arpi, jq is the matrix given by Lemma 7, that is:

Brp~ω iq Brp~ω jqi~

pI Arpi, jqqω.pi jq.Let

Zk 2pk 1q8k. (6.6)

Let us denote adW the operator H ÞÑ rW,Hs. The l.h.s. of (6.1) is then:

Hr V r 1

i~rHr,Wrs

8

j1

1

pi~qjj!adjWrpVrq

8

j2

1

pi~qjj!adjWrpHrq

that is

Hr V co,r V r V r V co,r 8

j1

1

pi~qjj!adjWrpVrq

8

j2

1

pi~qjj!adjWrpHrq.

orHr hr1 pVr V co,rq Vr R1 R2 (6.7)

Let us setVr1 : pVr V co,rq V r R1 R2. (6.8)

We want to estimate Vr1. We first prove the following proposition.

Proposition 17. Let W be in Jkpρq and 0 δ ρ. Then

rHr,W si~ρδ ¤ 1

eδpω Zk∇pB~

r B~0qρqW ρ. (6.9)

and for d ¥ 2,

1

d!rHr,W s, . . .loomoon

d times

spi~qdρδ ¤ δω

2πZkp1 Zk∇pB~

r B~0qρq

Zkδ2r

d

W dρ (6.10)

Let now Wr be the (scalar) solution of (6.3). Then, we have

Wrρ ¤ MMr

1 ZkDpB~r B~

0qρVrρ, (6.11)


and therefore for d ¥ 2,

1

d!rHr,Wrs, . . .loomoon

d times

spi~qdρδ ¤ ω1 ZkDpB~

r B~0qρ

2πZkδ

ZkMMrδ2r

1 ZkDpB~r B~

0qρVrρ

d

(6.12)(MM is defined in (1.3) and DBρ is meant for max

i1...m

°j1...m

∇iBjρ).

Proof. We first prove (6.9). Note that the proof is somehow close to the proof of Proposition16, items (1) and (2).

Since B0pLωq Lω, (5.12) reads

rH0,Wrsi~ρδ ¤ ω

eδWrρ. (6.13)

Note that prHr H0,Wrs~qij Grpω.i~qGrpω.j~q~ Wij where GrpY q B~

r pY q Y, Y P Rm

(note that Gr has an explicit dependence in ~ that we omit to avoid heaviness of notations).Indeed, since each Lωi is self-adjoint on L2pTlq, B~

r pLωq can be defined by the spectraltheorem. Hence, we have

rB~r pLωq,W sij pei, rB~

r pLωq,W sejq pei,B~r pLωqWej WB~

r pLωqeiq pB~

r pω.i~q B~r pω.j~qqpei,Wejq.

Using (5.13) we get that

rHr H0,Wrsi~ρδ ¤ Xrρδ. (6.14)

where Xr is defined through pXrqij Grpω.i~qGrpω.j~q~ pWrqij.

In order to estimate the norm of Xr we need to express its symbol Xr. This is donethanks to formula (5.7) and the fact that we know the matrix elements of Xr.Expressing pXrqij as a function of ppi jq~2, i jq through i, j ij

2 ij

2and using (5.7)

we get that

Xrpξ, q, ~q Grpω.ξ ω.q~2q Grpω.ξ ω.q~2q~

W 1rpω.ξ, q, ~q : X 1

rpω.ξ, q, ~q,

so that, using (remember that we denote p.ω.q °j1...m

°i1...l

pjωijqi)

»RmpGrpΞ ω.q~2q GrpΞ ω.q~2qqei Ξ,p¡dp 2 sin rp.ω.q~2s

»Rm

GrpΞqei Ξ,p¡dp,

X 1rpp, q, ~q

»Rm

Gri pp p1qsin rpp p1q.ω.q~2s~

W 1rpp1, q, ~qdp.

Therefore Xrρδ is equal to


m

i1

¸qPZl

»R2m

dpdp1|2k

γ1

µkγpp, qqBγ~Gri pp p1qsin rpp p1q.ω.q~2s

~Wrpp1, q, ~q

|epρδqpω|p||q|q

¤m

i1

¸qPZl

» k

γ1

µkγpp, qqγ

µ1

γµ

ν1

γ

µ

γ µ

ν

2|Bγµν~ Gri pp p1q||Bν~sin ppp p1q.ω.q~2q

~||Bµ~ Wrpp1, q, ~q|epρδqpω|p||q|qdpdp1 (6.15)

(6.16)

Using now the inequalities (5.21) and (5.22), we get,

|k

γ1

µkγpp, qqγ

µ1

γµ

ν1

γ

µ

γ µ

ν

Bγµν~ Gri pp p1qBν~

sin ppp p1q.ω.q~q~

Bµ~ Wrpp1, q, ~q|

¤ ω maxj1...m

|pjp1j||q|k

γ1

µkγpp, qqγ

µ1

γµ

ν1

γ

µ

γ µ

ν

|Bγµν~ Gri ppp1q||ppp1q.ω.q|ν |Bµ~ Wrpp1, q, ~q|.

Therefore we notice (after changing q Ø q1) that Xrρδ is majored by the maximum over~ P r0, 1s of

k

γ0

»R2m

¸pq,q1qPZ2l

µkγpp, qqω maxj1...m

|pj p1j|PpGri ,Wrqepρδqpω|p||q|qdpdp1 (6.17)

where P is defined in (5.27) and

Gri pΞ, xq Gri pΞq so thatGripp, qq Gri ppqδq0.

Therefore we can verbatim use the argument contained between formulas (5.27)-(5.29) andwe arrive, in analogy with (5.30), to the fact that p1 kq18kXrρδ is majored by themaximum over ~ P r0, 1s of

¸q,q1PZl

»R2m

ω maxj1...m

|pj||q1|k

γ,γ10

µkγpp, qq|Bγ~Gri pp, qq|µkγ1pp1, q1q|Bγ1~ Wrpp1, q1q|epρδqpω|p|ω|p1||q||q1|qdpdp1

¸q1PZl

»R2m

ω maxj1...m

|pj||q1|k

γ,γ10

µkγpp, 0q|Bγ~ Gri ppq|µkγ1pp1, q1q|Bγ1

~Wrpp1, q1q|epρδqpω|p|ω|p1||q1|qdpdp1

Since |Gri ppq||pj| |Gri ppqpj| |∇jGri ppq|, we get that (use ρδ ¤ ρ and again |q1|eδ|q1| ¤1eδ

)

Xrρδ ¤ p1 kq8kωeδ

∇GrρWrρ ¤ Zkω

eδ∇pB~

r B~0qρWrρ. (6.18)

Here ∇pB~r B~

0qρ is understood in the sense of (5.17)-(4.18).(6.9) follows form (6.13) and (6.14)-(6.18).


We will prove (6.10) by the same argument as in the proof of Proposition 16. Take(5.35) with G1 : 1

i~rWr, Hrs, Gs1 1i~rWr, Gss and γs ds

dδ for 1 ¤ s ¤ d 1, γ0 δ,

γd1 δd.

We get

1

d!Gdργ0 ¤ Zd1

k

d!e2pd1qγ0 γd2p δdqd1Wrd1

ρ G1ργd1

¤ Zd1k

d!d!e2pd1qp δdq2d2

Wrd1ρ G1ρδd

¤ 1 Zk∇pB~r B~

0qρd!d!e2d1p δ

dq2d1

Zd1k Wrdρ

¤ δ

2πd2e1Zkp1 Zk∇pB~

r B~0qρq

dded

?2πd

d!

2 Zkδ2

d

Wrdρ

and we get (6.10) by dded?

2πdd!

¤ 1 and d2e1 ¥ 1 if d ¥ 2, and setting ρ ρr, γ0 γr.

In order to prove (6.11) we first estimate V rρr defined by (6.5) where Arpi, jq is givenby (7.5).

Lemma 18. Let V 1 be defined by V 1ij Arpi, jqV co

ij . Then

V 1ρ ¤ ZkDpB~r B~

0qρV coρ. (6.19)

Proof. The proof will actually be close to the one of (6.9). Arpi, jqω.pijq Grpω.j~qGpωi~q~

so

Arpi, jq » 1

0

∇Grpp1 tqω.j~ tω.i~qdt.Therefore

V 1 1»

0

m

n1

∇nGrpp1 tqω.j~ tω.i~qV con dt

so

V 1 ¤ sup0¤t¤1

m

n1

∇nGrpp1 tqω.j~ tω.i~qV con .

Let Xrn be defined through

pXrnqij ∇nGrpp1 tqω.j~ tω.i~qpV co

n qij ∇nGrpω i j

2~ p1 2tqpi jq~

2qpV co

n qij

By the argument as before, using (5.7), we get that the symbol of Xr satisfies X rnpξ, qq

∇nGrpω ξ p1 2tqq ~2qVcon pξ, qq pX r

nq1pω ξ, qq : ∇nGrpω ξ p1 2tqq ~2qpVcon q1pω ξ, qq

since Vcon has the same structure as V so there exists pVcon q1 such that Vcon pξ, xq pVcon q1pω ξ, xq.


Taking now the Fourier transform of pX rnq1pΞ, qq with respect to Ξ one gets by translation-

convolution

pX rnq1pp, q, ~q

»Rm

∇nGrpp p1qeippp1q.ω.qp12tq~2 Vcon pp1, q, ~qdp1.So, as before,

|Bγ~pX r

nq1pp, q, ~q| ¤» γ

µ1

γµ

ν1

|Bγµν~ ∇nGrpp p1qBν~eippp1q.ω.qp12tq~2Bµ~ Vcon pp1, q, ~q|γµ

γ µ

ν

dp1

¤»Rm

γ

µ1

γµ

ν1

|Bγµν~ ∇nGrpp p1qp|p p1||q|2qνBµ~ Vcon pp1, q, ~q|γµ

γ µ

ν

dp1

Following the same lines than in the proof of (6.9) we get that (remember that, by definition,

Xrρ :m°n1

Xrnρ, Xr

nρ pX rnq1ρ by Definitions 11 and 10)

Xrρ ¤ Zk

m

i1

m

n1

∇nGri ρV con ρ ¤ Zk max

n∇nGrρ

m

n1

V con ρ ¤ Zk∇GrρV coρ

and the Lemma is proved.

Corollary 19. Let V 2 defined by V 2ij p1 Arpi, jqq1V co

ij . Then

V 2ρ ¤ 1

1 Zk∇pB~r B~

0qρV coρ. (6.20)

(6.11) is now a consequence of (5.15) and the fact that V coρ ¤ V ρ.(6.12) is obtained by putting (6.11) in (6.10). The proposition is proved.

We need finally the following obvious Lemma:

Lemma 20. Define

VMpx, ξq :¸

|q|¥MVqpξqeiqx. (6.21)

Then

VMρδ ¤ eδMVρ (6.22)

Corollary 21. Let V M be defined by

V Mij Vij when |i j| ¥M

0 when |i j| M.

Then

V Mρδ ¤ eδMV ρ. (6.23)


Proof. Just notice that the symbol of V M , VM , satisfies, by (5.7), VMpξ, q, ~q 0 when|q| ¤M and apply Lemma 20.

Let us define, for a decreasing positive sequence pρrqr0...8, ρr1 ρrδr to be specifiedlater,

Gr DpB~r B~

0qρr maxi1...m

¸j1...m

∇ipB~r B~

0qjρr . (6.24)

We are now in position to derive the following fundamental estimates of the five terms in(6.7):

hr1ρrδr ¤ hr1ρr V co,rρr Vrρr (6.25)

Vr V co,rρrδr V Mrρrδr ¤eMrδr

V rρr

Vr2

ρr (6.26)

V rρrδr ¤Zk

MMr

δ2r

p1 ZkGrq2 Vr2ρr (6.27)

R1ρrδr ¤Zk

MMr

δ2rp1ZkGrq1 Zk

MMr

δ2rp1ZkGrqV rρrVr2

ρr (6.28)

R2ρrδr ¤Zk

M2Mr

ωp1ZkGrqδ3rp1ZkGrq2

1 ZkMMr

δ2rp1ZkGrqV rρrVr2

ρr (6.29)

Indeed, p6.25q is obvious and p6.26q is nothing but Corollary 21.p6.27q is derived by using Proposition 16, item (5) equation (5.16), Lemma 19 and equation

(6.5). Note that, as pointed out before, V r is cut-offed as V co,r thanks to (3.6).p6.28q is obtained through the definition R1 °8

j11

pi~qjj!adjWrpVrq, the fact that, by

(5.11), 1pi~qjj!adjWr

pVrqρrδr ¤ pZkδ2rqjVrρrWrjρr and (6.11), so that

R1ρrδr ¤8

j1

pZkδ2rqjVrρr

MMr

1 ZkGr

Vrρrj

.

p6.29q is proven by the definition R2 8°j2

1pi~qjj!adjWr

pHrq and the fact that, by (6.12) we

have that 1pi~qjj!adjWr

pHrqρrδr ¤ ω 1ZkGrZkδr

ZkMMr δ2r

1ZkGr Vrρrj

.

Collecting all the preceding estimates together with the definition p6.8q :

Vr1 : pVr V co,rq V r R1 R2,

we obtain:


Proposition 22. For r 0, 1, . . . , we have

Vr1ρrδr ¤ FrVr2ρr eδrMrVrρr (6.30)

with

Fr MMrZkδ2rp1 ZkGrq2

1 p1 ZkGrq MMr

δrωp1 ZkGrq

1 MMr

1ZkGrZkδ2rVrρr

. (6.31)

7. Fundamental iterative estimates: Diophantine condition case

In all this section also the norm subscripts ω and k are omitted.Let 0 δ ρ. Let us recall that we want to find Wr such that

eiWr~ pHr VrqeiWr~ Hr1 Vr1 (7.1)

where Hr1 Hr hr1 and Hr B~r pLωq, hr1 Vr DrpLωq and

hr1ρ Vrρ ¤ Vrρ, Vr1ρδ ¤ DrVr2ρ. (7.2)

In the case where ω satisfies the Diophantine condition (1.4) we look at Wr solving:

1

i~rHr,Wrs Vr Vr V r (7.3)

with V r pV rl ql1...m given by

pV rl qij

prV rl , V

rlpijqsqij

i~ωlpijq pi jq , V rij : pI Arεpi, jqq1 pVrqij. (7.4)

Here Arpi, jq is the matrix given by Lemma 7, that is:

B~r p~ω iq Brp~ω jq

i~ pI Arpi, jqqω.pi jq, (7.5)

The l.h.s. of (7.1) is:

Hr Vr 1

i~rHr,Wrs

8

j1

1

pi~qjj!adjWrpVrq

8

j2

1

pi~qjj!adjWrpHrq

that is

Hr V r V r 8

j1

1

pi~qjj!adjWrpVrq

8

j2

1

pi~qjj!adjWrpHrq.

orHr hr1 V r R1 R2 (7.6)

Proposition 23. Let Wr the (scalar) solution of (7.3).Then, for d ¥ 2, 0 δ ρ 8,

1


d times

spi~qdρδ ¤ δω

2πZkp1 Zk∇pB~

r B~0qρq

Zkδ2

d

Wrdρδ (7.7)


Wrρδ ¤2τγp τ

eδqτ

1 ZkDpB~r B~

0qρVrρ, (7.8)

so

1


d times

spi~qdρδ ¤ ω1 ZkDpB~

r B~0qρ

2πZkδ

2τγp τeδqτ

1 ZkDpB~r B~

0qρVrρ

d

(7.9)(let us recall that MM is defined in (1.3) and DBρ is meant for max

j1...m

°i1...m

∇iBjρ).

Proof. The proof of Proposition 23 is the same than the one of Proposition 17 done indetails in Section 6. The only minor difference is the discussion of the small denominatorsand is adaptable without pain. We omit the details here.

Using notation p6.24q, Proposition 16, last item (5.10), and Proposition 17 we can derivethe following fundamental estimates of the four terms in (7.6)

hr1ρrδr ¤ hr1ρr Vrρr

V rρrδr ¤Zk

22τγp τeδr

qτδ2r

p1 ZkGrq2 Vr2ρr

R1ρrδr ¤Zk

γp τeδr

qτδ2rp1ZkGrq

1 Zkγp τeδr

qτδ2rp1ZkGrqV rρr

Vr2ρr

R2ρrδr ¤Zk

pγp τeδr

qτ q2ωp1ZkGrqδ3rp1ZkGrq2

1 Zkγp τeδr

qτδ2rp1ZkGrqV rρr

Vr2ρr

Collecting all the preceding results we get:

Proposition 24. For r 0, 1, . . . , Vr1ρrδr ¤ F 1rVr2

ρr with

F 1r

γp τeδrqτZk

δ2rp1 ZkGrq2

22τ p1 ZkGrq γ

δrp τeδrqτωp1 ZkGrq

1 γp τeδr

qτ1ZkGr

Zkδ2rVrρr

(7.10)

8. Strategy of the KAM iteration

In the case of the Diophantine condition, the strategy consists in finding a sequences δrsuch that, with F 1

r given by (7.10),8

r1

δr δ 8 andr¹i1

D2ri

i ¤ R2r , R ¡ 0. (8.1)

Indeed when (8.1) is satisfied and thanks to Proposition 24, the series8°r1

Vr, and therefore

8°r1

hr 8°r1

Vr are easily shown to be convergent in Jkpρ δ, ωq for ρ ¡ δ, at the condition


thatV ρ,ω,k R.

This last sum is the quantum Birkhoff normal form B~8 of the perturbation. Estimates on

the solution of the cohomological equations provide also the existence of a limit unitaryoperator conjugating the original Hamiltonian to its normal form.

The case of the Brjuno condition follows the same way, except that one has also to finda sequence of numbers Mr so that (6.31) holds. The main difference comes from the extralinear and non quadratic term in Proposition 22. This difficulty is overcome by derivingout of V ρ,ω,k a sequence of quantities with a quadratic growth as in (7.10). This leadsto an extra condition for the convergence of the iteration, condition involving only thearithmetical properties of ω and which can be removed by a scaling argument. These ideaswill be implemented in the following section.

9. Proof of the convergence of the KAM iteration

In this section the norm subscripts ω and k might be committed in the bodyof the proofs. They are nevertheless reestablished in the main statements.

This section is organized as follows: we first prove the convergence of the KAM iterationin the Brjuno case with a restriction on ω (Theorem 29), restriction released in Theorem30 thanks to the scaling argument already mentioned. This proves and precises Theorem1. We then prove the corresponding classical version (Corollary 35, global Hamiltonianversion of the singular integrability of [LS1]) leading to Theorem 2 precised. We end thesection by more refined results under Diophantine condition on ω, Theorem 39, leading tothe criterion contained in Theorem 3.

9.1. Convergence of the KAM iteration I: constraints on ω.

Proposition 25. Let us fix 0 C η 1, ρ ¡ 0 and let us choose

ρ0 ρ, δr α2r, 0 α ¤ log 2, and Mr 2r. (9.1)

For E ¥ E0 defined below by (9.13) let us suppose:8

r0

| logMMr |2r1

3log δr

2r logZkE

2r

Ck 8 (9.2)

and, for 1 ¤ r ¤ l,ZkGr η Cr, (9.3)

δ3r

δ3r1

epδr1Mr12δrMrq ¡ 2 (9.4)

andMMrZkδ2r

Vrρr 1

2p1 η Crq, (9.5)

together withZkG0 0, (9.6)

δ30

δ31

epδ1M12δ0M0q ¡ 2 (9.7)


andMM0Zkδ2

0

V ρ0 1

2. (9.8)

Then, for r ¥ 0

Vr1ρr1 ¤ pDkq2r1

. (9.9)

where

Dk : eCkpV ρ eαα3

2M21ZkE

q. (9.10)

Note that G0 0 and that, taking (9.3) for r 1 we get:

1 ¡ η ¡ Zk∇V 1ρ and C η Zk∇V 1ρ. (9.11)

Therefore we will impose the condition

∇V 1ρ η C

Zk(9.12)

Proof. We first prove the two following Lemmas.

Lemma 26. Under the hypothesis (6.30), (9.3) and (9.5), and η 1, we have that, if

E ¥ 3α p1 ηqωM1pωqp1 ηq2M1pωq : E0 (9.13)

then

Vr1ρr1 ¤ drVr2ρr eδrMrVrρr with dr

M2MrZkE

δ3r

.

The proof is immediate by noticing that, under proposition 22, (9.3) and (9.5), (6.31)gives that, for r 1, . . . ,

Fr ¤ MMrZkδ2rp1 η Crq2

3 2

MMr

δrωp1 η Crq

so, for r 0, 1, . . .

Fr ¤ MMrZkδ2rp1 ηq2

3 2

MMr

δrωp1 ηq

.

The case r ¥ 1 is obtained out of the preceding inequality, and the case r 0 comes formthe fact that ZkG0 0 ¤ η.

Therefore E must be ¥ 3ωp1ηqMMrδr

pωqp1ηq2 MMr

δrpωq

¤ 3ωp1ηqM1pωqαp1ηq2M1pωqα 3αωp1ηqM1pωq

p1ηq2M1pωq since MMr is

increasing with Mr and the Lemma is proved.

Lemma 27. Let Vr Vrρr eδrMr

2drwhere dr M2

MrZkE

δ3r, Vr satisfy (6.31) and V0 :

V, ρo ρ. Then

Vr1 ¤ drV2r . (9.14)


The proof reduces to completing the square in Proposition 22 and noticing that e2δrMr

4dr

eδr1Mr1

2dr1¡ 0 by (9.4), since MMr1 ¥ MMr . The Lemma has for consequence the fact

the

Vr1 ¤r¹s0

d2rs

s V 2r

0 ¤ peCk V0q2r . (9.15)

This concludes the proof of Proposition 25 since Vrρr ¤ Vr.

Proposition 28. Let Vrρr ¤ pDkq2r with Dk eP and Dk M , M and P definedbelow by (9.23) and (9.20). Then (9.3), (9.4) and (9.5) hold.

Note that8°r0

δr 2α.

Proof. (9.4):it is trivial to show that (9.4) is satisfied when α ¤ 2 log 2.(9.5):(9.5)-(9.8) are equivalent to

1

2rlogMMr

log δ2r

2r logZk

2r logDk 1

2rlog

1

2p1 η C

rq (9.16)

and

logM1 log δ20 logZk logDk log

1

2(9.17)

which is implied by

logDk 8

r0

| logMMr |2r

infr¥0

log δ2r logZk

2r ∆ (9.18)

8

r0

| logMMr |2r

logZk 2 logα 2

e ∆ (9.19)

which is implied by

logDk 8

r0

| logMMr |2r

logZk 2

e ∆ : P (9.20)

where

∆ infr¥1

"1

2rlog

1

2p1 η C

rq, log

1

2

* 8 for η 1. (9.21)

Note that ∆ ¡ 0, P ¡ 0.(9.3):

remember that Br1 Br V r1, and Br1ρr1 ¤ Brρr1 V 1r1ρr1 ¤ Brρr

V 1r1ρr . So DBr1ρr1 ¤ DBrρr1 DV 1

r1ρr1 .

Moreover one has, by (5.18), DV 1r1ρr1 δr

2¤ V 1r1ρr1

δr2e

¤ 2Vr1ρr1

δre¤ 2 pDkq2r1

δreout

of which we conclude that


ZkGr η Cr ùñ ZkGr1 η Cpr 1q, @r ¥ 1, if

2ZkD2r1

k

α2re C

r C

r 1 C

rpr 1q (9.22)

which is implied by Dk M for

M infr¥1

α2reC

2Zkrpr 1q2pr1q

αeC

8Zk

14 1 (9.23)

since α ¤ log 2, Zk ¥ 8 and C ¤ 1.

Proposition 28 together with Proposition 25 shows clearly that

(9.12) andDk eP and Dk M

ùñ Vrρr ¤ pDkq2r (9.24)

where Dk is given by (9.10) i.e. Dk : eCkV ρ eCk eδ0M0

2d0. Note that since M 1 so

is Dk ¤ M leading ot the superquadratic convergence of the sequence pVrqr0,.... In orderfor Dk to satisfy the two conditions of the bracket in the l.h.s. of (9.24) the two terms inDk will have to both satisfy the two conditions. This remark will be the key of the maintheorem below.

Let us denote by ωij, j 1 . . .m, i 1 . . . l be the ith component of the vector ωj. Letus remark that

M1pωq minj1...m

1

mini1...l

|ωij|and

1

M1pωq maxj1...m

mini1...l

|ωij|. (9.25)

Let us denote

Bpωq :8

r0

| logM2r |2r

. (9.26)

We have that, by (9.2) and (9.20),

Ckpωq 2Bpωq 6 logα 6 log 2 2 log pZkEq. (9.27)

and

P pωq Bpωq logZk 2

e ∆. (9.28)

Theorem 29. [Brjuno case] Let α, ρ, η, and C be strictly positive constants satisfying

α 2 log 2, ρ ¡ 2α, 0 C η 1. (9.29)

Let us define, for ∆ infr¥1

12r

log 12p1 η C

rq and M

αeC8Zk

14,

Rkpωq p1 ηq4M1pωq2p3α p1 ηqωM1pωqq2

α6e2Bpωq

26Z2k

min

"eBpωq∆

21eZk,M

*. (9.30)


Let us suppose that, in addition to Assumptions (A1), (A2) Brjuno case and (A3), ωsatisfies

3α p1 ηqωM1pωq2eαp1 ηq2M1pωq3 ¤ α3e2Bpωq

26Zkmin

"eBpωq∆

21eZk,M

*. (9.31)

and the perturbation V satisfies

V ρ,ω,k Rkpωq, ∇V 1ρ,ω,k η C

Zk. (9.32)

Then the BNF as constructed in section 2 converges in the space J :k pρ 2α, ωq to B~

8and

B~8 B~

0ρ2α,ω,k OpV 2ρ,ω,kq as V ρ,ω,k Ñ 0. (9.33)

That is to say that there exists a (scalar) unitary operator U8 such that the family ofoperators H pHiqi1...m, Hi Lωi Vi, satisfies, @~ P p0, 1s,

U18 HU8 B~

8pLωq. (9.34)

U8 is the limit as r Ñ 8 of the sequence of operators Ur eiWr~ . . . ei

W0~ constructed in

Section 2 and

U8 UrBpL2pTlqq ¤Ar~ O

E2r

~

as r Ñ 8 for some E 1,

here Ar is defined by (9.53).Moreover, U8 I P J~

0 pρ 2α, ωq and

U8 I~ρ2α,ω,0 O

V ρ,ω,0~

as V ρ,ω,0 Ñ 0, (9.35)

and, for any operator X for which there exists Xk,ρ such that for all W P Jkpρ, ωq,

rX,W si~ρδ,ω,k ¤ Zkδ2Xk,ρW ρ,ω,k, (9.36)

U18 XU8 X P Jkpρ 2α δ, ωq and

U18 XU8 Xρ2αδ,ω,k ¤ D

δ2sup

ρ2α¤ρ1¤ρXk,ρ1 O

V ρ,ω,kδ2

supρ2α¤ρ1¤ρ

Xk,ρ1

(9.37)

where D is given by (9.57).

Note that the second condition in (9.32) “touches” only the average V and not the full

perturbation V . It can also be replaced for any ρ1 ¡ ρ by V ρ1 ¤ V ρ1 ¤ eρ1ρZk

since

∇V 1ρδ ¤ V 1ρeδ

for any δ ¡ 0.


9.2. Convergence of the KAM iteration II: general ω. Before we start the proof oftheorem 29, let us show the way of overcoming the condition (9.31).We first notice that multiplying the family LωV by λ ¡ 0 preserves of course integrability.Moreover λpLω V q Lλω λV .On the other side we see easily that:

Bpλωq Bpωq2 log λ, M1pλωq λ1M1pωq and therefore ωM1 is invariant by scaling.(9.38)

Let us show that, for λ large enough, (9.31) will be satisfied for ωλ : λω. More precisely,let us define

µ α 2rp1 ηqα p1 ηqωM1pωqs2eαp1 ηq2M1pωq3

26Zkα3e2Bpωq

ν eBpωq∆

21eZk

we easily see that the following number λ0 is uniquely defined:

λ0 λ0pωq : inf tλ ¡ 0 such that Mλ µ ¥ 0 and νλ3 µ ¥ 0u sup

"µ

M,µν

13

*.

(9.39)Elementary algebra leads to

Lemma. @ω, @λ ¥ λopωq, (9.31) is satisfied for ωλ : λω.Since the BNF of λH is the BNF of H multiplied by λ we get that the latter will exist

and be convergent if λV ρ,λω,k ¤ Rkpλωq and λ∇V 1ρ,λω,k ¤ ληCZk

. we get the

Theorem 30. Let α, ρ, η, and C be strictly positive constants satisfying

α 2 log 2, ρ ¡ 2α, 0 C η 1. (9.40)

Let us define ∆ infr¥1

12r

log 12p1 η C

rq, M

αeC8Zk

14 inf

r¥1

α2reC

2Zkrpr1q

2pr1q

and,

for λ ¥ λ0pωq given by (9.39),

Rλ,kpωq Rkpλωqλ

λp1 ηq4M1pωq2

p3α p1 ηqωM1pωqq2α6e2Bpωq

26Z2k

min

"λ2 e

Bpωq∆

21eZk,M

*. (9.41)

Let us suppose that the general assumption (A1), (A2) Brjuno case and (A3) hold and

V ρ,λω,k Rλ,kpωq, ∇V 1ρ,λω,k ¤ η C

Zk. (9.42)

Then the BNF as constructed in section 2 converges in the space J :k pρ 2α, λωq to B~

8and

B~8 B~

0ρ2α,λω,k OpV 2ρ,λω,kq. (9.43)


That is to say that there exists a (scalar) unitary operator U8, U8 I P Jkpρ 2α, λωqsuch that the family of operators H pHiqi1...m, Hi Lωi Vi, satisfies, @~ P p0, 1s,

U18 HU8 B~

8pLωq. (9.44)

U8 is the limit as r Ñ 8 of the sequence of operators Ur eiWr~ . . . ei

W0~ constructed in

Section 2 and

U8 UrBpL2pTlqq ¤Ar~ O

E2r

~

as r Ñ 8 for some E 1,

here Ar is defined by (9.53).Moreover, U8 I P J~

0 pρ 2α, λωq and

U8 I~ρ2α,λω,0 O

V ρ,λω,0~

as V ρ,λω,0 Ñ 0, (9.45)

and, for any operator X for which there exists Xk,ρ,λ such that for all W P Jkpρ, λωq,rX,W si~ρδ,λω,k ¤ Zk

δ2Xk,ρ,λW ρ,λω,k, (9.46)

U18 XU8 X P Jkpρ 2α δ, λωq and

U18 XU8 Xρ2αδ,λω,k ¤ D

δ2sup

ρ2α¤ρ1¤ρXk,ρ1,λ O

V ρ,λω,kδ2

supρ2α¤ρ1¤ρ

Xk,ρ1,λ

(9.47)

where D is given by (9.57).

Remark 31. Note that, for λ large enough, Rλ,kpωq λRkpωq. Therefore the radius ofconvergence increases by dilating ω. But this fact is compensated by the fact that thenorm in the condition of convergence (9.42) (we take here V 0), V 1ρ,λω,0 Rλ,kpωq,increases at least as λ (an actually highly non sharp estimate as the Gaussian case showsclearly) when λ is large, as shown by the following lemma. Therefore the optimization onλ of (9.42) remains between bounded values of λ.

Lemma 32. For ω1 ¥ ω, Fρ,ω1,k Fρ,ω.k ¥ pω1 ωqρ∇Fρ,ω,k.The proof is an immediate consequence of

eρ1X eρX eρXpepρ1ρqX 1q ¥ eρXpρ1 ρqX, X ¥ 0.

9.3. Proof of Theorem 29. First notice that B~rB~

r1 V r Vrp, 0, ~q so B~rB~

0ρr ¤r°1

Vll which is convergent under (9.31) and (9.32). What is left is to show that the

sequence of unitary operators Ur : eiWr~ . . . ei

W1~ converges to a unitary operator on L2pTlq.

This is done by proving that the sequence Ur is Cauchy (~ P p0, 1s). For p ¡ n let us denote

Enp eiWnp

~ eiWnp1

~ . . . eiWn1

~ I, (9.48)

so that Unp Un EnpUn. We have for all r,

eiWr~ I Tr with Tr i

Wr

~

» 1

0

eitWr~ dt. (9.49)


Therefore

~TrBpL2pTlqq ¤ WrBpL2pTlqq ¤ Wrρr,k. (9.50)

By (6.11) we have also that

Wrρr Wrρr,k ¤ MMr

1 η Cr Vrρr,k ¤MMr

1 η CrD2r

k l ¡ 0

W0ρ W0ρ,k ¤ M1V ρ,k (9.51)

Note that, by the Brjuno condition, we have for all rMMr ¤ eBpωq2r

and, by the conditionon Dk insuring the convergence of the BNF, Dk eBpωq, so that:

A :8

r1

MMr

1 η CrD2r

k ¤ peBpωqDkq2r1 η Cr 8. (9.52)

We also define, for n ¥ 1,

An 8

rn

MMr

1 η CrD2r

k . (9.53)

Note that An OpE2nq as nÑ 8 for E eBDk 1 by (9.24).By (9.49) we get that

Enp eiWnp

~ Enp1 I eiWnp

~

eiWnp

~ Enp1 Tnp

eiWnp

~ eiWnp2

~ Enp1 eiWnp

~ Tnp1 Tnp. (9.54)

By iteration we find easily that

Enp p

k2

eiWnp

~ . . . eiWnpk1

~ Tnpk eiWnp

~ Tnp1 Tnp

and, by unitarity of eiWr~ and (9.50),

EnpBpL2pTlqq ¤p

k0

TnkBpL2pTlqq ¤p

k0

Tnkρnk ¤p

k0

Wnkρnk~

¤8

k0

Wnkρnk~

¤ An~Ñ 0 as nÑ 8 since A 8.

So EnpBpL2pTlqq Ñ 0 as n Ñ 8 and so does Unp Un EnpUn by unitarity of Un, andUn converges to U8 in the operator topology. Moreover we get as a by-product of thepreceding estimate that

U8 UrBpL2pTlqq ¤Ar~.

Since U8 is a perturbation of the identity which doesn’t belong to any Jpρq, ρ ¡ 0,there is no hope to estimate U8ρ,ω,k. Nevertheless, and somehow more interesting, we


will estimate U8 I in the ρ2α,ω,0 topology. In the sequel of this proof we will denote ρ : ρ,ω,0 and use the fact that ρr ¥ ρ2α, @r P N.

We first remark that, for r ¥ 0, since ~ρ ¤ ρ,

Tr~ρ2α eiWr~ I~ρ2α ¤ eiWr~ I~ρr ¤8

j1

pZ0qj1Wr

~ jρrj!

eZ0Wrρr

~ 1

Z0

We first remark also thatpI Tr1qUr Ur1.

Therefore, denoting Pr Ur I,

Pr1 pI Tr1qPr Tr1

so

Pr1ρ ¤ PrpZ0Tr1ρ 1q Tr1ρ pPrρ 1

Z0

qpZ0Tr1ρ 1q 1

Z0

so Pr1ρ 1Z0¤ pPrρ 1

Z0qpZ0Tr1ρ 1q and

Pr1~ρ2α 1

Z0

¤ pP0~ρ 1

Z0

qr1¹j1

pZ0Tj~ρj 1q ¤ pP0~ρ 1

Z0

qr1¹j1

eWjρj

~

e

r1°j1

Wjρj~ pP0~ρ

1

Z0

q

¤ eA~ pP0~ρ

1

Z0

qTherefore

U8 I~ρ2α P8~ρ2α ¤ eA~

M1V ρ

~ 1

Z 0

1

Z 0

Let us note that, by construction, A Op D2k

1η q and that Dk depends on η through (9.10).

Lemma 33. Dη ηpV ρq such that

D2k

1 η OpV ρq as V ρ Ñ 0.

Proof. By looking at the expression of the radius of convergence which tends to 0 as η Ñ 1we see that as V Ñ 0 one can take values of η Ñ 1 which makes the second term in thedefinition of Dk of order V ρ and the ratio Dk

1η of order V ρ.

By application of the Lemma we find that

U8 I~ρ2α P8 ¤ eA~

M1V ρ

~ 1

Z 0

1

Z 0 O

V ρ~

.

which gives (9.45).

In order to prove (9.47) we first denote Vr eiWr~ .


We have, actually for any operator X, that VrXV1r X

8°j1

1j!

adjWrpXq.

Let us suppose now that DXρ,k such that for all W P Jkpρq

rX,W si~ρδ ¤ Zkδ2Xk,ρW ρ. (9.55)

(e.g. Xk,ρ Xk,ρ).Using (5.11) we get (since 2πl ¡ 1q we get

VrFV 1r ρδ ¤ F ρ

1 Zkδ2Wrρ

, (9.56)

and also (let us recall again that ρr1 ρr δr, ρ0 ρ)

1

j!~jadjWr

pXqρrδrδρr1δ ¤Zkδ2r

j1

rX,Wrsi~ρrδWlj1ρrδ

¤Z0

δ2r

j1

rX,Wrsi~ρrδWlj1ρr ,

out of which we get

V0XV1

0 Xρ1δ ¤ rX,W0si~ρ0δ1 Zk

δ20W0ρ

.

by V1V0XV1

0 V 11 V1XV

11 V1pV0XV

10 XqV 1

1 and (9.56)

V1V0XV1

0 V 11 V1XV

11 ρ2δ ¤ V0XV

10 Xρ1δ

1 Zkδ21W1ρ1

¤ rX,W0si~ρ0δp1 Zk

δ20W0ρqp1 Zk

δ21W1ρ1q

and by iteration

UrXU1r UrV

10 XV0U

1r ρr1δ ¤ rX,W0si~ρ0δ

p1 Zkδ20W0ρq . . . p1 Zk

δ2rWrρrq

and by X Ñ V 10 XV0

UrV 10 XV0U

1r UrV

11 V 1

0 XV0V1U1r ρr1δ ¤ rX,W1si~ρ1δ

p1 Zkδ21W1ρq . . . p1 Zk

δ2rWrρrq

VrXV 1r Xρr1δ ¤ rX,Wrsi~ρrδ

1 Zkδ2rWrρr

.


So that by summing the telescopic sequence we get

UrXU1r Xρr1δ ¤

r

s0

rX,Wssi~ρsδr¹js

1

1 Zkδ2jWjρj

¤r

s0

Xk,ρs

δ2ZkWsρse

r°js

log p1Zkδ2j

Wjρj q

and since p1 aqp1 2aq ¥ 1 if 0 a ¤ 12

UrXU1r Xρr1δ ¤

r

s0

Xk,ρs

δ2ZkWsρse

r°js

log p12Zkδ2j

Wjρj q

¤r

s0

Xk,ρs

δ2ZkWsρse

2r°js

Zkδ2j

Wjρj

¤r

s0

Xk,ρs

δ2ZkWsρse

28°j0

Zkδ2j

Wjρj

¤sup

ρ2α¤ρ1¤ρXk,ρ1

δ2Zk

r

s0

Wsρse2B

¤sup


δ2D

with B Zkα2M1V ρ

8°j1

Zk2jMMj

α2p1ηCjqD2j

k 8 and

D ZkpM1V ρ Aqe2B OpV ρq (9.57)

by Lemma 33.Therefore we get, by letting r Ñ 8 so that ρr Ñ ρ 2α,,

U18 XU8 Xρ2αδ ¤ D

δ2sup

ρ2α¤ρ1¤ρXk,ρ1 . (9.58)

8 ¡ U18 XU8 Xρ2αδ O

V ρδ2

supρ2α¤ρ1¤ρ

Xk,ρ1

(9.59)

The theorem is proved.

Remark 34. [Diophantine case] In the Diophantine case one immediately sees that

Bpωq ¤ 2 log rγ2τ s (9.60)

Moreover one easily sees that Rkpωq and Rλ,kpωq, together with λ0pωq, are decreasingfunctions of Bpωq. Therefore Rkpωq ¥ RDio

k pωq and Rλ,kpωq ¥ RDioλ,k pωq where RDio

k pωq and

RDioλ,k pωq are obtain by replacing Bpωq by 2 log rγ2τ s in the r.h.s. of (9.30) and (9.41). It


follows that Theorem 29 (resp. Theorem 30) is valid with Rdiok pωq in place of Rkpωq (resp.

Rdioλ,kpωq in place of Rλ,kpωq).

9.4. Convergence of the KAM iteration III: the classical limit. Since all the es-timates are uniform in ~, the methods of the present paper allow to prove the followingresult

Corollary 35. Let H a family of m ¤ l classical Hamiltonians pHiqi1...m on T pTlq of theform Hpx, ξq ω ξ Vpx, ξq H0pω ξq V 1pω ξ, xq. Then, under the hypothesis on ωof Theorem 30 (resp. Theorem 29) and the conditions

tHi,Hju 0 1 ¤ i, j ¤ m

Vρ Rλ,0pωq presp. R0pωqq∇V 1ρ η C

Z0

H is (globally) symplectomorphically and holomorphically conjugated to B08pω.ξq: for all

δ ¡ 0,there exist a symplectomorphism Φ1

8 such that

H Φ18 B0

8pH0q.Moreover, Φ1

8 I P Jpρ 2α δq (in particular Φ18 is holomorphic) and for any positive

δ ρ,

Φ18 Iρ2αδ ¤ D

δVρ (9.61)

where D is given by (9.68) below.Finally, for any function X satisfying (9.66), we have

X Φ18 X ρ2αδ ¤ D

δ2sup

ρ2α¤ρ1¤ρX 0,ρ1 O

V ρδ2

supρ2α¤ρ1¤ρ

X 0,ρ1

(9.62)

where X 0,ρ1 is defined in (9.66).

Proof. Once again the function B~8 is by construction uniform in ~ P r0, 1s so it has a limit

B08 as ~ Ñ 0. It is easy to get convinced that the construction of B0

8 is the same as theone of B~

8 after the substitution (we use capital letters for operators and calligraphic onesfor their symbols at ~ 0):

AB ÝÑ A BrA,Bsi~

ÝÑ tA,BueiW~ ÝÑ e LW

eiW1~ ei

W2~ ÝÑ e LW1 e LW2

eiW~ Aei

W~ ÝÑ A e LW .

Here is the usual function multiplication, t., .u denotes the Poisson bracket and e LW theHamiltonian flow at time 1 of Hamiltonian W (Lie exponential).


What is left is to prove the convergence of the sequence of flows e LWr . . . e LW1 : Φr asr Ñ 8. This is done by the same Cauchy argument than in the proof of Theorem 29.

For Φ : T Tl Ñ T Tl we denote Φρ 2l°i1

Φiρ where Φi are the components of Φ and

we define Enp by

Enp e LWnp e LWnp1 e LWn1 ITTlÑTTl ,

so that Φnp Φn Enp Φn.We will need the following

Lemma 36. Let Fpz, θq be analytic in t|=z|, |=θ| ¤ ρu. Then,

FL8p|=z|,|=θ|¤ρq ¤ Fρ. (9.63)

Proof. As in section 5 write

|Fpz, θq| |¸q

» Fpp, qqei p,ξ¡i q,x¡dp| ¤¸q

»| Fpp, qq|eρp|p||q|qdp F ρ.

We will denote

8ρ L8p|=z|,|=θ|¤ρq.Proposition 37. Let Hρ tpz, θq. |=z| ¤ ρ and |=θ| ¤ ρu.

Under the hypothesis of Theorems 29 and 30, Φr is analytic Hρ Ñ Hρr .

Remember that ρr ρr1°j0

δr, δr α2r.

Proof. Let us first remark that the “rule” rA,Bsi~ ÝÑ tA,Bu is in fact (and of course) a

Lemma.

Lemma 38. Let F P Jmpρq, G P J1pρq. Then rF,Gsi~ is the Weyl quantization of a function

σ~ on T Tl and

lim~Ñ0

σ~ σ0 tF ,Gu.

The proof is an easy exercise which consists (again) in computing the symbol of rF,Gsi~

through its matrix elements using Proposition 15, after expressing these matrix elementsout of the ones of F,G, themselves expressed through the symbols F ,G of F,G thanksof formula (5.7). The limit ~ Ñ 0 leads naturally to the Poisson bracket. Since thesetechniques have been extensively used through the present article, we omit the details.

Let us, by a slight abuse of notation, define again adW the operator F ÞÑ tW ,Fu.Being uniform in ~, the formula (5.11) taken with k 0 leads, for any F P Jmpρrq, to

1

j!adjWr

pFqρrδr ¤Z0

δ2r

j

FρrWrjρr


Let us denote ϕr e LWr and adWrpFq : tWr,Fu. Since F ϕ1r

8°j0

1j!

adjWrpFq we get

F ϕ1r ρr1 ¤

Fρr1 Z0

δ2rWrρr

.

Therefore, under the hypothesis of Theorem 30 (resp. Theorem 29), F ϕ1r is analytic in

Hρr1 for all F analytic in Hρr and so ϕ1r maps analytically Hρr1 to Hρr and so ϕr maps

analytically Hρr to Hρr1 . Writing Φr ϕr ϕr1 ϕ1 gives the result.

Let us write now for all r in N,

ϕr I Tr,with, as for (5.18),

Tr8ρ2α ¤ Tr8ρrδr 2l

i1

pTrqi8ρrδr ¤ ∇Wr8ρrδr : maxi

¸j

p∇jWrqi8ρrδr ¤Wr8ρreδr

,

and so

∇Tr8ρrδr ¤∇Wr8ρrδr2

eδr2 ¤ 4Wr8ρre2δ2

r

¤ Wr8ρrδ2r

.

In analogy with (9.54) we write

Enp ϕnp pEnp1 Iq I ϕnp pEnp1 Iq ϕnp pϕnp Iqso

Enp8ρ2α ¤ ∇ϕnp8ρ2αEnp18ρ2α Tnp8ρ2α

and, by induction,

Enp8ρ2α ¤p1

k0

Tnk8ρ2α

p1¹sk

p∇ϕns8ρ2αq Tnp8ρ2α

¤p1

k0

Tnk8ρ2α

p1¹sk

p1 ∇Tns8ρ2αq Tnp8ρ2α

¤p

k0

Tnk8ρ2α

8¹sk

p1 ∇Tns8ρ2αq

¤p

k0

Wnkρnkδnk

e

8°s0

Wsρsδ2s

¤ AneA Ñ 0 as nÑ 8.

Here we defined

A :8

j1

MMl

δ2l p1 η ClqD

2l

k 8, (9.64)


An 8

jn

MMl

δlp1 η ClqD2l

k Ñ 0 as nÑ 8 since A 8 (9.65)

and used (9.51).So Φr converges to Φ8 in the L8pHρq topology.The proof of (9.62) is exactly the one of (9.47) by using the dictionary expressed earlier.

Since it is a by-product of the proof of (9.61) we repeat it here. We denote ρ : ρ,ωand use the fact that ρr ¥ ρ2αδ, @r P N.

We have, actually for any operator X , that X ϕ1r X

8°j1

1j!

adjWrpX q.

Taking (9.55) at k 0 we have

tX ,Wuρδ ¤ Z0

δ2X 0,ρWρ (9.66)

We have

F ϕ1r ρδ ¤ Fρ

1 Z0

δ2Wrρ

, (9.67)

and also (let us recall again that ρr1 ρr δr, ρ0 ρ)

1

j!~jadjWr

pX qρrδrδρr1δ ¤Z0

δ2r

j1

tX ,Wrui~ρrδWrj1ρrδ

¤Z0

δ2r

j1

tX ,Wrui~ρrδWrj1ρr ,

out of which we get

X ϕ10 X ρ1δ ¤ tX ,W0uρ0δ

1 Z0

δ20W0ρ

.

by X ϕ10 ϕ1

1 X ϕ11 pX ϕ1

0 X q ϕ11 and (9.67)

X ϕ10 ϕ1

1 X ϕ11 ρ2δ ¤ X ϕ1

0 X ρ1δ1 Z0

δ21W1ρ1

¤ tX ,W0uρ0δp1 Z0

δ20W0ρqp1 Z0

δ21W1ρ1q

and by iteration

X Φ1s X ϕ0 Φ1

s ρs1δ ¤ tX ,W0uρ0δp1 Z0

δ20W0ρq . . . p1 Z0

δ2sWsρsq


By the same argument we get

X ϕ0 Φ1s X ϕ0 ϕ1 Φ1

s ρs1δ ¤ tX ,W1uρ1δp1 Z0

δ21W1ρq . . . p1 Z0

δ2sWsρsq

X ϕ1r X ρr1δ ¤ tX ,Wruρrδ

1 Z0

δ2rWrρr

.

so that by summing the telescopic sequence

¤r

s0

Xk,ρs

δ2Z0Wsρse

r°js

log

1Z0

δ2j

Wjρj

and since p1 aqp1 2aq ¥ 1 if 0 a ¤ 12

¤r

s0

Xk,ρs

δ2Z0Wsρse

r°js

log

1 2Z0

δ2j

Wjρj

¤r

s0

Xk,ρs

δ2Z0Wsρse

2r°js

Z0δ2j

Wjρj

¤r

s0

Xk,ρs

δ2Z0Wsρse

28°j0

Z0δ2j

Wjρj ¤sup


δ2Z0

r

s0

Wsρse2B

¤sup


δ2D

with B Z0

α2M1V ρ 8°j1

Z02jMMj

α2p1ηCjqD2j

k 8 and

D Z0pM1V ρ Aqe2B OpV ρq (9.68)

by Lemma 33.Therefore we get, by letting r Ñ 8 so that ρr Ñ ρ 2α,

X Φ18 X ρ2αδ ¤ D

δ2sup

ρ2α¤ρ1¤ρX k,ρ1 O

V ρδ2

supρ2α¤ρ1¤ρ

X k,ρ1

. (9.69)

Let now X P tξ1, . . . , ξl, x1, . . . , xlu, tX ,W1u BΞW1 where Ξ is the conjugate quantityto X . Therefore tX ,W0uρδ ¤ ∇W0ρδ ¤ 1

δW0ρ. Therefore X k,ρ 1 and X Φ1

r X ρr1δ ¤ D

δ2, @X P tξ1, . . . , ξl, x1, . . . , xlu which means that

Φ1r Iρr1δ ¤

Dδ2

(9.70)

In fact we just proved that Φ18 I Φ with Φρ8δρ2αδ ¤ D

δ2. Corollary 35 is

proved.


9.5. Convergence of the KAM iteration IV: the Diophantine case. Though theDiophantime case is covered by the Theorem 29 (see Remark 34), we can also use directlyProposition 24 in order to low down the hypothesis of the Theorem.

In fact Proposition 24 shows that the same proof will be possible by only replacingMMrpωq by γp τ

eδrqτ and (9.13) by

E ¥ 22τα 2rα p1 ηqωγp τeαqτ s

p1 ηq2γp τeαqτ E1 (9.71)

Indeed (7.10) is verbatim the same as (6.31) after replacing MMrpωq by γp τeδrqτ and the

first term in the parenthesis, namely 1, by 22τ . Therefore the proof will be the same byreplacing Bpωq by Bαpγ, τq

Bαpγ, τq 8

r0

log pγp τeδr

qτ q2r 2 logγp τeαqτ 2τ log 2 2 log

2τγp τ

eαqτ

and of course Ck and P by the corresponding expressions C 1k, P

1.The very last change will concern Dk which now will be Dk eCkV ρ because the

estimate of Proposition 24 reads now directlyV r1ρlδl ¤ F 1rV r2

ρl: this will imply that

in the Diophantine case there is no condition for ω similar to (9.31), and no conditionα 2 log 2. We get:

Theorem 39. [Diophantine case] Let α, ρ, η, C and E be strictly positive constants satis-fying

ρ ¡ 2α, 0 C η 1. (9.72)

Let us define, for ∆ infr¥1

12r

log 12p1 η C

rq and M

αeC8Zk

14,

Rkpωq p1 ηq2γp τ

eαqτ

22τα 2rα p1 ηqωγp τeαqτ s

2α6

26Z2kp2τγp τeαqτ q4

min

"p2τγp τeαqτ q2e∆

21eZk,M

*.

(9.73)Then if

V ρ,ω,k Rkpωq, ∇V 1ρ,ω,k η C

Zk, (9.74)

the same conclusions as in Theorem 29 and Corollary 35 hold.

9.6. Bound on the Brjuno constant insuring integrability. As mentioned in theintroduction we can use Theorem 30 to estimate the rate of divergence of the Brjunoconstant as the system remains integrable while the perturbation is vanishing.

Let us suppose that we let ω vary in a way such that ω remain in a bounded set rω, ωsof p0,8q. (9.73) tells us that, in order that Theorem 39 holds, ω can be taken as we wantas soon as as γ and τ satisfy V ρ,ω,k r.h.s. of (9.73) and ∇Vρ,ω,k ηC

Zk. It is easy


to check that, since Bαpγ, τq : 2 log2τγp τ

eαqτ Ñ 8 as γ, τ or both of them diverge, we

have, for Bαpγ, τq large enough (in order that the min in (9.73) is reached by the first termand that 22τ Bαpγ, τq),

Rkpωq ¥ 2Ke3Bαpγ,τq

with K p1ηq4α6

pα2p1ηqωq2621eZ3k. Therefore for ∇V 1ρ,ω,k ηC

Zkand V ρ,ω,k small enough

(namely V ρ,ω,k ¤ 2Ke3Bα where B

α is the smallest value of Bαpγ, τq which makes themin in (9.73) reached by the first term and which is larger than 22τ ), we have

Corollary 40. The conclusions of Theorem 39 hold as soon as

Bαpγ, τq 1

3log

1

V ρ,ω,k

1

2log 2K.

Remark. In the case of the Brjuno condition, Theorem 29, it happens that λ0pωq C 1e2Bpωq and Rλ0pωq C as Bpωq Ñ 8 for some bounded constants C,C 1. Therefore ourcondition of convergence takes the form V ρ,C1e2Bpωqω,k C. This leads to a sufficientcondition on Bpωq depending on the way V Ñ 0. For example it is easy to check that,if V Ñ 0 as V εV0, ε Ñ 0 and V0 with a symbol V 1

0 whose Fourier transform in ξ iscompactly supported, one gets a condition of the form Bpωq D log log 1

ε D1 for some

constants D, D1.

10. The case m l

Lemma 41. Let the vectors ωj, j 1 . . .m l, be independent over R and let Ω the

matrix of matrix elements Ω pΩijqi,j1...l with Ωij : ωji . Then

(1) any V satisfies (1.5)(2) @q P Zl, q 0, min

1¤i¤m|xωi, qy|1 ¤ l|Ω| (there is no small denominator).

Proof. Ω is invertible by the independence of the ωjs. This proves (1). Moreover one hasimmediately that 1 ¤ |q| ¤ l|Ω1| max

j1...l|xωj, qy|.

Therefore the main assumption reduces to:Main assumptions (extreme case)

ωj P Rl, j 1 . . . l, are independent over R and rHi, Hjs 0, @1 ¤ i, j ¤ l. (10.1)

References

[Ar2] V.I.Arnold Geometric Methods in the Theory of Ordinary Differential Equations, Springer-Verlag (1985)[BGP] D. Bambusi, S.Graffi, T.Paul Normal Forms and Quantization Formulae, Comm.Math.Phys. 207, 173-195

(1999)[MB] Max Born, The mechanics of the atom, Ungar Pub. Co. (1925).[Bru] A.D. Brjuno. Analytical form of differential equations. Trans. Mosc. Math. Soc, 25,131-288(1971); 26,199-

239(1972), 1971-1972.[CdV1] Y. Colin de Verdi6re, Spectre conjoint d’oprateurs pseudo-diffrentiels qui commutent, I-Le cas non int-

grable, Duke Math. J. 46, 169-182 (1979)


[CdV2] Y. Colin de Verdi6re, Spectre conjoint d’oprateurs pseudo-diffrentiels qui commutent, II-Le cas intgrable,Math. Z. 171, 51-73 (1980).

[CdV3] Y.Colin de Verdiere, Quasi-modes sur les varietes Riemanniennes, Inventiones Mathematicae 43, 129-159(1981)

[De] A.Deprit, Canonical transformations depending on a small parameter, Cel.Mech.Dyn.Astr. 1, 12-30 (1969)[El] L. H. Eliasson, Normal forms for Hamiltonian systems with Poisson commuting integrals—elliptic case,

Comment. Math. Helv., 65(1),4–35,1990[FM] F.M.Fedoryuk, V.P.Maslov, Semi-classical approximation in quantum mechanics D.Reidel, (1981)[Fo] G.Folland, Harmonic Analysis in Phase Space, Annals of Mathematics Studies 122, Princeton University

Press (1989)[Ga] G.Gallavotti, Criterion of Integrability for Perturbed Nonresonant Harmonic Oscillators, ”Wick Ordering”

of the Perturbations in Classical Mechanics and Invariance of the Frequency Spectrum, Comm.Math.Phys.87, 365-383 (1982)

[GG] S. Graffi, V. Grecchi, and B. Simon, Borel summability: application to the anharmonic oscillator, Phys.Lett. B 32, 631-634 (1970)

[GP] S. Graffi, T.Paul, Convergence of a quantum normal form and an exact quantization formula, Journ. Func.Analysis, 262, 3340-3393 (2012).

[Ho] G.I.Hori, Theory of general perturbations with unspecified canonical variables, Publ. Astr. Soc. Japan 18,287-296 (1966)

[It] H. Ito. Convergence of birkhoff normal forms for integrable systems. Comment. Math. Helv., 64:412–461,1989.

[TK] T. Kato, Perturbation theory of linear operators, Springer (1988).[KP] S. Kuksin and G. Perelman, Vey theorem in infinite dimensions and its application to KdV, Discrete and

Continuous Dynamical Systems. Series A 27, 1-24 (2010)[PU] T. Paul and A. Uribe, A construction of quasimodes using coherent states, Annales I.H.P., Physique

Theorique 59, 357-381 (1993)[HP] H. Poincare, Les methodes nouvelles de la mecanique celeste, T. 2 Gauthier-Villars (1892).[Po1] G.Popov, Invariant Tori, Effective Stability, and Quasimodes with Exponentially Small Error Terms I -

Birkhoff Normal Forms, Ann.Henri Poincare 1, 223-248 (200)[Po2] G.Popov, Invariant Tori, Effective Stability, and Quasimodes with Exponentially Small Error Terms II -

Quantum Birkhoff Normal Forms, Ann.Henri Poincare 1, 223-248 (2000)

[Ru] H.Russmann, Uber die Normalform analytischer Hamiltonscher Differentialgleichungen in der Nahe einergleichgewichtlosung, Math.Annalen 169, 55-72 (1967)

[Ra] J. V. Ralston, On the construction of quasimodes associated with stable periodic orbits, Commun. Math.Phys. 51, 219-242 (1976)

[BS] B. Simon, Determination of eigenvalues by divergent perturbation series, Adv. in Math. 7, 240-253 (1971)[LS1] L. Stolovitch, Singular complete integrabilty, Publ. Math. I.H.E.S., 91, 133-210, (2000)[LS] L.Stolovitch, Progress in normal form theory, Nonlinearity 22, 7423-7450 (2009)[LS2] L. Stolovitch. Normalisation holomorphe d’algebres de type Cartan de champs de vecteurs holomorphes

singuliers. Ann. of Math., (161):589–612, 2005.[JV1] J. Vey, Sur certains systmes dynamiques sparables, Am. Journal of Math. 100 591614, (1978)[JV2] J. Vey, Algbres commutatives de champs de vecteurs isochores, Bull. Soc. Math. France,107 423432, (1979)[Zu] N.T.Zung, Convergence Versus Integrability in Normal Form Theory, Ann. of Math.161, 141-156 (2005)[WW] E. T. Whittaker and G. N. Watson, A course of modern analysis. An introduction to the general theory

of infinite processes and of analytic functions: with an account of the principal transcendental functions,Fourth edition, Cambridge University Press, 1927.

CNRS and Centre de Mathematiques Laurent Schwartz, Ecole Polytechnique, 91128 PalaiseauCedex France

E-mail address: [email protected]

CNRS and Laboratoire J.-A. Dieudonne Universite de Nice - Sophia Antipolis Parc Valrose 06108Nice Cedex 02 France

E-mail address: [email protected]

quantum singular complete integrability · initiated in classical mechanics in [de, ho] and uses...

Documents